Towards 5G-enabled vehicular communication systems, message dissemination in a mobile, low-latency environment, like public safety, is critical to the timely and reliable delivery of emergency messages. It is well-known that classical cryptographic primitives such as RSA and ECC are no longer secure (feasible) in the future when QC is practical. Proposition 1 In the spirit of such observations, this paper provides a lightweight post-quantum authentication scheme for emergency message propagation in vehicular fog networks. The proposed technique relies on CRYSTALSDilithium, a NIST-standard post-quantum lattice-based digital signature scheme, to achieve quantum-safe authentication at low cost. The protocol has four phases: system initialization, vehicle emergency token generation, fog-level validation, and distributed revocation. It adopts a semi-trusted architecture consisting of certified On-Board Units (OBUs), roadside fog nodes, and a central authority. Empirical results show a 2.65 ms signing time, 1.43 ms verification time, and 3.2 KB token size, all leading to prototyping in 5G network real-time applications. Experimental results reveal >30% reduction of computational delay with lower communication overhead compared with other existing schemes and without sacrificing security. The results demonstrate that the proposed scheme achieves post-quantum security, scalable revocation, and the preservation of privacy in fog computing environments that are characterised by limited computing capability. This paper presents a practical, future-proof authentication model well-suited for emergency response in next-generation vehicular networks.

Optimising the control of wind power systems based on Doubly-Fed Induction Generators (DFIG) raises complex technical challenges, intrinsically linked to the non-linear natureof these machines. With this in mind, this study presents a comparison of two distinct control approaches: Proportional-Integral (PI) control, and Backstepping control, designed specifically to address the challenges posed by unstable and variable dynamics.The methodological approach is based on a DFIG model built on the foundations of vector control. This theoretical framework is implemented into a MATLAB/Simulink environment. Backstepping control, in particular, is stabilised by means of a rigorous construction of the Lyapunov function, guaranteeing error convergence and robustness in the face of disturbances. The simulation results highlight the differences in performance. Whilethe classic PI control approach is robust to parametric variations, it results in a slower response time (27.6 ms) and higher static error (0.2%). Its simple structure and efficient implementation make it a reliable choice in industrial environments with limitedresources. In contrast, the Backstepping method significantly reduces overshoot, improves system response time (0.18 ms), and achieves a notable reduction in static error (0.064%), demonstrating its superiority in dynamic and unstable environments. This approach excels in managing the non-linearities inherent in wind energy systems, giving it a clear advantage in unstable or fluctuating environments. In short, this study does not simply juxtapose two methods; it outlines the future of more adaptive, more responsive control. While PI remains a faithful ally in simplicity, Backstepping technology offers a promising approachto the development of smart energy systems.

Autonomous Underwater Vehicles (AUVs) face challenges in maintaining stability during visual Simultaneous Localization and Mapping (vSLAM) operations, particularly when affected by internal solitary waves (ISWs). This study presents a novel integrated control strategy that combines a Nonlinear Disturbance Observer (NDO) and Nonlinear Model Predictive Control (NMPC), specifically adapted to address the nonlinear and unpredictable nature of ISW disturbances in underwater environments. Unlike previous NDO-NMPC implementations in other domains, this framework incorporates dynamic modeling of ISWs and underwater-specific tuning mechanisms to maintain robust vSLAM performance. To the best of our knowledge, this is the first study integrating NDO with NMPC for AUV-vSLAM under ISW disturbances. The proposed method is evaluated using a custom AUV model integrated with the ORB-SLAM2 framework, tested through Software-in-the-Loop (SITL) simulations under various ISW intensities. Results show that the NDO-NMPC algorithm outperforms traditional PID, Sliding Mode Control (SMC), and standalone NMPC controllers in terms of stability, trajectory tracking, and mapping accuracy. This approach reduces the impact of ISWs, improves the number of visual feature points for mapping, and achieves lower Root Mean Square Error (RMSE) in position and velocity. This work offers a robust solution for improving AUV navigation and mapping in dynamic underwater environments, with potential applications in autonomous underwater exploration and surveying.

Water quality and availability require continuous, timely, and cost-effective monitoring solutions. This article presents a systematic review of literature published between 2020 and 2025 on sensors and the Internet of Things (IoT) for intelligent water monitoring, focusing on measurement parameters, communication architectures, and analytical approaches. The research contribution is the development of a structured taxonomy of sensors and analytes, a comparative analysis of network technologies and management platforms, and a synthesis of real-world applications across diverse domains with field validation. The methodology followed predefined inclusion and exclusion criteria, systematic variable extraction, and the PRISMA flow diagram, ensuring traceability and reproducibility, while validation indicators such as accuracy, drift, packet loss, and autonomy were normalized for comparison. The results highlight a convergence toward a physicochemical core of parameters (pH, temperature, turbidity, conductivity/total dissolved solids, and oxidation–reduction potential) and the predominance of LoRa/LoRaWAN integrated with platforms such as The Things Network (TTN) and ThingSpeak, combined with protocols including Message Queuing Telemetry Transport (MQTT), Node-RED, and Blynk. Applications include photovoltaic buoys for marine environments, rural networks with limited infrastructure, and UAV–LoRa relays extending coverage up to 10 km. Advances in wastewater treatment incorporate soft sensors and optical techniques, while analytics such as edge computing, machine learning, and federated learning demonstrate improved accuracy and timeliness of decisions. This review concludes with design and standardization guidelines addressing sensor–network–energy–validation integration and outlines future research on scalability, robustness, cybersecurity, and interoperability. The findings provide practical guidance for sustainable water management.

This study addresses the challenge of long-term peak load forecasting in power systems by enhancing Convolutional Neural Networks (CNN) with an efficient two-stage hyperparameter optimization strategy. The research contribution is a sequential Random–Bayesian approach that first explores the hyperparameter space broadly using Random Search and then refines the search with Bayesian Optimization, improving prediction accuracy while controlling computational cost. Experiments on daily peak load data from New South Wales (Australia), spanning 2015–2022, demonstrate that the proposed two-stage strategy achieves the lowest mean absolute error (MAE = 695.57), mean absolute percentage error (MAPE = 7.39%), and competitive root mean square error (RMSE = 900.68) compared to single-stage Random Search and Bayesian Optimization. Although the two-stage method requires higher training time than Random Search, it offers a balanced trade-off between accuracy and computational efficiency. These findings provide a practical reference for designing robust hyperparameter optimization strategies in CNN-based long-term load forecasting, thereby supporting more informed decision-making for grid reliability and capacity planning.

Managing concurrent high‑resolution video from multiple UAVs is challenging because prior systems often optimize transmission or perception in isolation and rarely validate real‑time, multi‑stream operation with operator tooling and fault recovery. Research contributions are: a centralized, thread‑per‑stream architecture with a 3‑frame drop‑oldest capture buffer and on‑device post‑processing that sustains real‑time throughput while keeping the GUI responsive; standardized multi‑object‑tracking reporting with MOTChallenge metrics; quantified autonomous reconnection with low variance and disclosure of resource usage and reproducible engineering levers. Each RTSP feed is handled by a dedicated worker (capture→decode→infer→render), detection uses YOLOv8n with temporal identity assignment and inference is standardized (640×640, conf 0.5, NMS IoU 0.5). Experiments run on a laptop‑class platform with 720p streams. The system sustains 29.46 FPS (SD 0.50; 95% CI [29.36–29.56]) for a single stream and 28.56 FPS (SD 0.65; 95% CI [28.43–28.69]) for three concurrent streams (≈3.1% drop), with utilization increasing from CPU 25.86%→50.24% and GPU 21.73%→35.76% yet remaining below saturation. Recovery is 6.16 s after UAV reboot and 3.51 s after camera disconnect, without disrupting other feeds. Under MOTChallenge (IoU 0.5), daytime tracking attains MOTA 0.84, IDF1 0.87, MT 92% / ML 8%; partial occlusion yields MOTA 0.73, IDF1 0.70, MT 78% / ML 22%. These results demonstrate real-time, multi-stream operation with quantified robustness. The current scope is limited to one and three streams with a single YOLOv8n backbone, and these constraints motivate future scaling to five or more concurrent streams as well as comparative evaluations across YOLOv8 variants and higher-capacity YOLO models.

In the rapidly evolving field of neuromarketing, understanding the neural basis of consumer decision-making is crucial. Identifying neural signals associated with buying intention can provide valuable insights into consumer behaviour, enhancing the effectiveness of marketing strategies. This study investigates buying intention by analyzing EEG signals recorded from 28 participants while they viewed video advertisements. The EEG data were acquired using the OpenBCI system, with electrodes placed at Fp1, Fp2, F7, F8, O1, and O2. The data was preprocessed to remove artifacts, segmented into, and decomposed into alpha, beta, and gamma frequency bands. Time-domain features (mean, mean absolute value, standard deviation, and Hjorth parameters) and frequency-domain features (power spectral density) were extracted. The total number of features obtained was 126 features. In order to reduce redundancy and to avoid overfitting in machine learning algorithm, feature selection was applied using the Pearson correlation approach. The result showed that distinct neural activity patterns were observed between high and low buying intention states, with the gamma band showing the most pronounced differences. Specifically, high buying intention conditions exhibited dominant activity across both frontal and occipital regions. Furthermore, classification analysis using machine learning, with feature selection, showed that the Random Forest algorithm achieved the highest accuracy of 98.7% with 58 features. Despite the limitations of a small sample size and a sparse 6-electrode configuration, the model achieved 94.4% accuracy using just six optimally selected features. These findings not only highlight the significance of the gamma band in understanding the buying intention states but also suggest practical implications for neuromarketing applications.

Modern Connected and automated vehicles expand the attack surface across in-vehicle networks and V2X links, making robust, evidence-based protection essential. This article presents a systematic review of vehicle cybersecurity, consolidating methods, datasets, and deployment barriers to inform research and practice. The research contribution is a focused synthesis (2020-2025) that (i) maps dominant technical themes (intrusion detection, secure in-vehicle networking, V2X protection, secure OTA, privacy), (ii) catalogs datasets and evaluation practices, and (iii) identifies reproducibility and real-world validation gaps that constrain deployment. Following PRISMA, we searched Scopus, IEEE Xplore, Web of Science, and MDPI within 2020-2025 using predefined Boolean strings. Inclusion targeted vehicle-focused cybersecurity methods or systems with empirical evaluation (simulation, testbed, or field); exclusions removed non-vehicular, purely conceptual, or duplicate records. Data extraction covered threat models, algorithm families, datasets, metrics, and deployment context; study quality and risk of bias were appraised using a standardized rubric. Results show progress in ML-based intrusion detection, cryptography and key management, authentication/access control, secure OTA updates, and V2X security; however, heterogeneous datasets and metrics limit cross-study comparability, and evaluations remain predominantly offline or lab-based. Bridging lab-to-road requires standardized datasets and metrics, multi-site and on-road evaluations, and co-assurance of security with safety and privacy requirements; we outline an actionable roadmap to accelerate deployment while mitigating methodological and publication biases.

Overtourism has emerged as a critical issue in popular tourist destinations, often leading to environmental strain, reduced visitor satisfaction, and safety concerns. Traditional methods such as ticket counts, or vehicle estimation fail to provide real-time insights or adapt effectively to dynamic outdoor environments. This study proposes a privacy-aware, real-time visitor capacity monitoring system for smart tourism, utilizing YOLOv8-based head detection and Centroid Tracking to ensure accurate, non-intrusive people counting in dense and complex crowd scenarios. Head detection is employed specifically to preserve personal privacy without compromising on detection performance. The system was trained on a custom dataset comprising over 3,000 annotated frames with diverse lighting conditions, occlusion levels, and viewing angles. Deployment at Wana Wisata Kawah Putih, an open-air tourist destination in Indonesia, demonstrated strong performance with 94.2% accuracy, 95.1% precision, and 90.6% recall, while sustaining >60 FPS for real-time execution. The integration of Centroid Tracking enables lightweight, frame-to-frame identity association with minimal computational overhead, making the system suitable for deployment on moderate-performance hardware. Despite its robustness, the system's performance slightly degrades under extreme weather (e.g., fog, direct glare) and rapid lighting transitions, which remain challenges for visual models. Moreover, the current model requires further evaluation for cross-location generalizability. Future research will explore the integration of predictive analytics for visitor flow forecasting, and further optimization of energy efficiency and adaptive detection under environmental uncertainty. This work contributes a scalable, ethical solution for real-time crowd monitoring to support informed, sustainable tourism management.

The integration of eye-gaze technology into robotic control systems has shown considerable promise in enhancing human–robot interaction, particularly for individuals with physical disabilities. This study investigates the influence of eye morphology and the use of corrective eyewear on the spatial accuracy of gaze-based robot control under static head pose conditions. Experiments were conducted using advanced eye-tracking systems and multiple machine learning algorithms—decision tree, support vector machine, discriminant analysis, naïve bayes, and K-nearest neighbor—on a participant pool with varied eye shapes and eyewear usage. The experimental design accounted for potential sources of bias, including lighting variability, participant fatigue, and calibration procedures. Statistical analyses revealed no significant differences in gaze estimation accuracy across eye shapes or eyewear status. However, a consistent pattern emerged: participants with non-monolid eye shapes achieved, on average, approximately 1% higher accuracy than those with monolid eye shapes—a difference that, while statistically insignificant, warrants further exploration. The findings suggest that gaze-based robotic control systems can operate reliably across diverse user groups and hold strong potential for use in assistive technologies targeting individuals with limited mobility, including those with severe motor impairments such as head paralysis. To further enhance the inclusiveness and robustness of such systems, future research should explore additional anatomical variations and environmental conditions that may influence gaze estimation accuracy.