This research is focused on traffic congestion in the small island of Malta which is the most densely populated country in the EU with about 1,672 inhabitants per square kilometre (4,331 inhabitants/sq mi). Furthermore, Malta has a rapid vehicle growth. Based on our research, the number of vehicles increased by around 11,000 in a little more than 6 months, which shows how important it is to have an accurate and comprehensive means of collecting data to tackle the issue of fluctuating traffic in Malta. In this paper, we first present the newly built comprehensive traffic dataset, called MalTra. This dataset includes realistic trips made by members of the public across the island over a period of 200 days. We then describe the methodology we adopted to generate syntactic data to complete our data set as much as possible. In our research, we consider both MalTra and the Q-Traffic dataset, which has been used in several other research studies. The statistical ARIMA model and two graph neural networks, the spatial temporal graph convolutional network (STGCN) and the diffusion convolutional recurrent network (DCRNN) were used to analyse and compare the results with existing research. From the evaluation, we found that the DCRNN model outperforms the STGCN with the former resulting in MAE of 3.98 (6.65 in the case of the latter) and a RMSE of 7.78 (against 12.73 of the latter).

The P versus NP problem is a cornerstone of theoretical computer science, asking whether problems that are easy to check are also easy to solve. "Easy" here means solvable in polynomial time, where the computation time grows proportionally to the input size. While this problem's origins can be traced to John Nash's 1955 letter, its formalization is credited to Stephen Cook and Leonid Levin. Despite decades of research, a definitive answer remains elusive. Central to this question is the concept of NP-completeness. If even one NP-complete problem could be solved efficiently, it would imply that all problems in NP could be solved efficiently, proving P equals NP. This research proposes that a notoriously difficult NP-complete problem can be solved efficiently, thereby potentially establishing the equivalence of P and NP.

Functional near-infrared spectroscopy (fNIRS) is a preferred neuroimaging technique for studies requiring high ecological validity, allowing participants greater freedom of movement. Despite its relative robustness against motion artifacts (MAs) compared to traditional neuroimaging methods, fNIRS still faces challenges in managing and correcting these artifacts. Many existing MA correction algorithms notably lack validation on real data with ground-truth movement information. In this work, we combine computer vision, ground-truth movement data, and fNIRS signals to preliminarily characterize the association between specific head movements and MAs. Fifteen participants (age = 22.27 ± 2.62 years) took part in a whole-head fNIRS study, performing controlled head movements along three main rotational axes. Movements were categorized by axis (vertical, frontal, sagittal), speed (fast, slow), and type (half, full, repeated rotation). Experimental sessions were video recorded and analyzed frame-by-frame using the SynergyNet deep neural network to compute head orientation angles. Maximal movement amplitude and speed were extracted from head orientation data, while spikes and baseline shifts were identified in the fNIRS signals. Results showed that head orientation and movement metrics extracted via computer vision closely aligned with participant instructions. Additionally, repeated and Up/Down movements tended to compromise fNIRS signal quality. The occipital and pre-occipital regions were particularly susceptible to MAs following Up/Down movements, while temporal regions were most affected by bendLeft/bendRight and Left/Right movements. These findings underscore the importance of cap adherence and fit in the relationship between movements and MAs. Overall, the work lays the foundation for an automated approach to developing and validating fNIRS MA correction algorithms.

Scaling massive multiuser multiple-input multipleoutput (MIMO) to larger antenna numbers constrains each chain in terms of power consumption and implementation. In this paper, the use of low resolution digital-to-analog converters (DACs) enabling these large scale arrays in downlink multiuser massive MIMO is studied with a focus on reaching adjacent channel leakage ratio (ACLR) constraints. A quantization aware precoder with constraints on out-of-band (OOB) emissions is designed. The proposed precoder is also studied in the light of power amplifier (PA) efficiency through peak to average power ratio (PAPR). Using conventional DACs, a small BER degradation and low PAPR is ensued, but the ACLR target is missed by a large margin. With finite impulse response (FIR)-DACs, the ACLR target is reached at the expense of greater hardware requirements and some PAPR regrowth.

Machine learning (ML) has proven to be a key enabler of various industrial cyber-physical systems-empowered functionalities, such as predictive maintenance (PdM) in Industry 4.0. Among many ML methods, the scope of this paper is self-organizing map (SOM). SOM has many attractive properties for industry applications (e.g., unsupervised learning, noise robustness). Further, it can yield various tasks (e.g., clustering, dimensionality reduction, health indicator construction, visualization, etc.), rendering it versatile for data-driven PdM, including anomaly detection, diagnosis, and prognosis tasks. This papers presents a brief review of the main SOM applications in the PdM field, as well as future research opportunities.

We explore the use of Large Language Models (LLMs) to enable the dynamic creation of decentralized infrastructure and intelligence at the edge through AI-powered chatbots. Leveraging prompt engineering and the extensive knowledge embedded in LLMs, we explore how models such as OpenAI GPT-4o and GPT-3.5-Turbo can assist users in assembling and interconnecting nearby discovered devices for distributed computing and sensing tasks. We create a dataset of smartphone specifications detailing device resource characteristics, which is used to analyze optimal recommendations for infrastructure formation. This is validated through human inspection to assess the decision-making process of the LLMs. Furthermore, we explore the potential of open-source LLMs (LLaMa2) to create scalable AI-chatbots, which could drive broader adoption of the proposed method. Our findings highlight both the current strengths and limitations of LLMs, pointing to key challenges and opportunities for fully exploiting their advanced reasoning capabilities. Our research paves the way to simplify the deployment of decentralized intelligence at the edge, offering insights into the future of AI-driven, user-guided infrastructure formation.

The expansion of transformer-based architectures has introduced challenges in balancing computational efficiency with task-specific performance, particularly as models grow in scale and complexity. Contextual gradient pruning offers a novel solution through the dynamic selection of parameters based on gradient magnitudes, ensuring that critical linguistic and semantic elements are retained while reducing computational overhead. The methodology introduces a sparsification objective function that aligns parameter reduction with evolving training dynamics, enabling adaptability without compromising representational capacity. Experiments conducted on a state-of-the-art open-source language model demonstrate the method's ability to maintain robust performance across perplexity, token prediction accuracy, and downstream task benchmarks. Results reveal that computational efficiency improves significantly, with reductions in floating-point operations exceeding 60% at higher sparsity levels, while linguistic coherence and rare word coverage remain largely preserved. Comparative analysis highlights the superiority of gradient-informed pruning over conventional static approaches, showcasing its scalability and robustness across diverse architectures. Scalability is further evidenced through consistent improvements in inference latency, energy consumption, and model transferability, making the approach suitable for both small-scale and resource-intensive deployments. The analysis of adversarial robustness and attention head utilization provides deeper insights into the model's adaptive capabilities under varying sparsity conditions. Additionally, the method demonstrates potential in enabling broader accessibility to language models through its ability to tailor sparsity configurations to specific computational environments. The findings position contextual gradient pruning as a transformative advancement in efficient language model optimization, addressing key challenges associated with sustainability and scalability. By combining theoretical rigor with practical efficacy, the study contributes a significant step forward in the pursuit of more adaptable and resourceconscious natural language processing systems.

Dynamic wireless charging (DWC) of electric vehicles (EVs) can greatly alleviate "short driving range" concern, making it a highly feasible prospect for electrified transportation. To achieve maximum power transfer in DWC, it is essential that the primary coil (under the road) and secondary coil (below the EV) are fully aligned. Lateral misalignment (LTM) occurs when the coils fail to align properly. The energy losses caused by LTM can be compensated using conventional controllers, however, they cannot perform well in higher misalignment scenarios and exhibit overshooting, and steady-state error (SSE). In this paper, a model prediction-based approach is implemented to design a controller that nullifies the LTM effect and compensates for the reduced power transfer to the EV. This work presents a comparative study between the model predictive controller (MPC) and conventional Proportional-Integral (PI) controller in terms of performance and suitability. The MPC controller uses an optimization algorithm to select the optimal control action over the prediction horizon, minimizing SSE and reducing overshooting in the system, thereby performing exceptionally well across all degrees of LTMs. The effectiveness of the proposed MPC has been evaluated through simulations in MATLAB/Simulink and experimental tests. The proposed MPC demonstrates superior performance compared to the PI controller in both simulation and experimental evaluations.

This paper investigates the computation of the radar cross section (RCS) of equilateral triangular targets using Monte Carlo numerical integration. The analysis employs the Physical Theory of Diffraction (PTD) to model the scattering behavior and numerically evaluates the RCS for a canonical planar geometry. A Python-based implementation is provided for an equilateral triangle with side lengths of 2 meters, illuminated by a 10 GHz radar wave. The simulation yields an RCS value of approximately 0.0292 m², showcasing the effectiveness of the Monte Carlo approach for RCS estimation in simple geometries. This work contributes to the understanding of electromagnetic scattering and stealth design methodologies.

Recent advancements in wireless technologies, particularly in the context of the sixth generation (6G) mobile communications and Internet of Things (IoT) systems, have introduced a wide range of requirements and challenges in wireless communication. These developments necessitate comprehensive channel information, encompassing the three-dimensional (3D) features of electromagnetic (EM) signals. Such signals are now typically transmitted by antenna arrays, manipulated by Reconfigurable Intelligent Surface (RIS) structures, and received by another set of antenna arrays. This complex propagation environment demands sophisticated modeling techniques to accurately capture and predict channel behavior, essential for the design and optimization of next-generation wireless systems. Ray tracing (RT) based simulators have gained significant traction in recent years, proving their worth in accurately and efficiently simulating EM propagation environments. These simulators have demonstrated remarkable accuracy in modeling complex wireless scenarios, making them invaluable tools for researchers and engineers. However, the intricate interactions between EM waves and the physical environment in three-dimensional space can still render the simulation process time-consuming, especially for scenarios with high complexity. Fortunately, recent state-of-the-art developments in Graphics Processing Unit (GPU) and Central Processing Unit (CPU) technologies have substantially mitigated this challenge. These hardware advancements have dramatically reduced the computational overhead associated with complex environmental simulations, making it feasible to conduct comprehensive and realistic RT-based analyses of sophisticated wireless environments. This paper introduces WiPy-RT, a high-performance ray tracing simulator designed for modeling RIS-enabled wireless environments. WiPy-RT features an interactive interface that leverages highly configurable and open-source Python libraries for 3D structures, including PyVista, Trimesh, and PyTorch3D. This GPUaccelerated simulator excels in simulating first-order diffraction and multi-reflection (up to 6 orders of reflection) ray tracing scenarios, while being capable of modeling RIS behavior with large-scale unit cell arrays. WiPy-RT's fast execution and ability to handle complex 3D environments make it particularly suitable for advanced wireless communication research, including the simulation of massive MIMO systems and intelligent reflecting surfaces.

The Hierarchical Entropy-Based Analysis (HEBA) framework introduces an innovative approach to ransomware detection through the application of advanced entropy metrics and hierarchical data examination. By employing multifaceted entropy measures, including Shannon entropy, conditional entropy, and hybrid metrics, HEBA effectively identifies anomalous patterns indicative of malicious encryption activities. The framework's probabilistic assessment of deviations from established benign profiles enhances detection precision, thereby minimizing false positives. Comprehensive evaluations demonstrate HEBA's high detection accuracy and low false positive rates across diverse ransomware variants. Resource utilization assessments indicate minimal impact on system performance, affirming the framework's operational efficiency. Detection latency measurements reveal prompt identification of ransomware activities, enabling timely intervention. The analysis of entropic patterns provides deeper insights into ransomware encryption behaviors, enhancing threat intelligence. Comparative analyses highlight HEBA's superiority over traditional detection methods, showing its potential to advance cybersecurity defenses. The findings contribute to the broader understanding of entropy-based detection mechanisms and their practical applications in combating sophisticated cyber threats.

The rise of eHealth technologies has transformed cardiac disease diagnosis, leveraging edge computing, AI, and IoT to offer critical insights into heart health. Data privacy constraints in centralized systems hinder access to large-scale ECG datasets, posing challenges for early diagnosis. While advancements in quantization and compression enable neural networks to run on edge devices, effective solutions for efficient training and inference on constrained devices are limited. To address challenges to training on the edge, we propose a mix-precision quantized DNN FPGA accelerator designed for multi-class cardiac diagnosis. Our solution achieves a top-1 test accuracy of up to 93.26% while enhancing computational efficiency, optimizing resource usage, and reducing transmission power. Our Mix-Precision Quantized FPGA Accelerator achieves up to 136x and 7.2x faster inference compared to state-of-theart Split-CNN and DCNN-Convolutional FPGA Accelerators, respectively. The accelerators offer a throughput of up to 1439.36 samples per second, latency of only 695µs, and programming logic power consumption staying below 600mW. Using hardwaresoftware co-design, our FPGA-based "Training on the Edge" approach combines software flexibility with hardware speed and improved diagnostic Top-1 test accuracy by up to 2.8% within just five training cycles, making the model more diverse to the dataset. The proposed approach also accelerates the development and reduces hardware rebuild time by a factor of (Training Cycles-1)x, ensuring efficient, sustainable ML solutions on edge devices. The source code is made available on https://github.com/shakeelakram00/Continual-Learningon-FPGAs-using-FINN

HVDC technology has long been recognized for its efficiency in high-capacity, long-distance power transmission. Unlike conventional AC systems, HVDC allows for the direct transmission of electricity without the losses typically associated with long-distance AC transmission. The ability to control power flows across large networks with minimal loss, while also facilitating the integration of renewable energy, makes HVDC an ideal solution for modern power grids.As global energy demands continue to grow and as more renewable energy sources are integrated, there is a pressing need to enhance the control mechanisms of HVDC systems. The traditional methods for controlling HVDC systems, which often rely on manually configured parameters and linear models, are no longer sufficient to handle the complexities of contemporary grids. This is where Artificial Intelligence (AI) comes into play. AI can improve the flexibility, speed, and responsiveness of HVDC control systems, optimizing their performance in real-time and ensuring that they meet the evolving demands of modern power grids.

Electrical systems, both at the level of local microgrids and larger, interconnected grids, have always required sophisticated control and protection systems to ensure safe, efficient, and reliable operations. Traditionally, these systems have relied on manually programmed algorithms and physical components that react to faults or instability. However, with the advent of AI and machine learning, there is a growing shift toward autonomous, intelligent systems that can adapt in real-time to changing conditions. This shift has the potential to revolutionize the management of electrical grids, providing not only more efficient operation but also more effective protection mechanisms that can prevent catastrophic failures before they happen.At the heart of any electrical grid or microgrid system is a sophisticated control mechanism. This mechanism ensures that electricity is generated, transmitted, and distributed in a manner that maintains a stable frequency and voltage, optimizes load balancing, and meets real-time demand. Additionally, protection systems are essential to detect faults or abnormal conditions in the grid, isolating affected areas to prevent further damage. AI’s role in both control and protection systems is poised to enhance these capabilities, making systems more predictive, responsive, and resilient to both expected and unexpected challenges[1-37].

As humanity advances toward permanent lunar exploration and habitation, the establishment of reliable, autonomous power systems becomes a cornerstone for the success of lunar missions. Lunar microgrids, which are responsible for supplying power to habitats and critical infrastructure on the Moon, must operate with a high degree of autonomy due to the unique challenges of the lunar environment. One critical aspect of lunar power system operation is frequency regulation. Unlike Earth’s power grids, which benefit from a wide range of dynamic grid support mechanisms, lunar microgrids must manage frequency regulation autonomously due to the absence of external support and the extreme conditions on the Moon. This paper explores the importance of autonomous frequency regulation in lunar microgrids, discussing the challenges of maintaining grid stability, ensuring energy efficiency, and mitigating power fluctuations in a microgrid system. Additionally, the role of autonomous control systems, energy storage technologies, and advanced predictive algorithms in achieving effective frequency regulation on the Moon is analyzed. By examining the operational and technical requirements for lunar energy systems, this paper highlights how autonomous frequency regulation is crucial for ensuring reliable and continuous power supply for long-term lunar habitation and exploration. Through a combination of AI-driven optimization techniques, real-time monitoring, and predictive maintenance, lunar microgrids can achieve the resilience and stability required to meet the energy demands of lunar habitats, rovers, and other essential systems.

As humanity looks toward the Moon for future exploration and potential habitation, energy management becomes one of the most critical challenges to ensure the success of long-term lunar missions. Lunar microgrids, designed to power lunar habitats and scientific equipment, must operate autonomously due to the extreme remoteness and harsh environmental conditions on the Moon. The integration of autonomous energy management systems (AEMS) can significantly improve the efficiency, resilience, and sustainability of lunar microgrids. This paper explores the essential role of autonomous energy management in lunar microgrids, emphasizing the challenges of lunar power generation, energy storage, and distribution. The unique environmental conditions on the Moon, including extended lunar nights, temperature extremes, radiation, and dust, demand a high level of system autonomy to guarantee continuous power supply without human intervention. Autonomous systems can automate power balancing, optimize energy use, and ensure the smooth operation of microgrids even during emergencies or failures. This paper also discusses current terrestrial applications of autonomous energy management, their adaptation for lunar use, and the integration of advanced technologies such as artificial intelligence (AI), machine learning, and predictive algorithms. Finally, the paper presents case studies and simulations illustrating the potential benefits of autonomous energy management for lunar microgrids and highlights areas for further research and development.

Autonomous control systems (ACS) are crucial for the efficient operation and long-term sustainability of lunar microgrids, which are designed to power habitats, life support systems, and scientific equipment on the Moon. As lunar missions transition from short-term visits to permanent habitation, microgrids must become self-sustaining and resilient to ensure continuous energy supply, even in the face of unpredictable challenges like extended lunar nights, solar radiation, dust accumulation, and system malfunctions. Autonomous systems can significantly enhance the management of these grids by automating fault detection, power distribution, load balancing, and system reconfiguration without requiring human intervention. This paper reviews the importance of ACS in lunar microgrids, focusing on the key features and benefits of such systems. It discusses the unique challenges of lunar environments, including radiation, temperature fluctuations, and the lunar dust problem, and examines the potential of ACS to address these challenges. Moreover, it explores current terrestrial autonomous control solutions, the adaptation of these technologies for lunar applications, and the integration of advanced machine learning algorithms for predictive maintenance and adaptive control. The paper also presents a case study analyzing the feasibility of implementing ACS in lunar microgrids. The review concludes by identifying critical areas for further research, ensuring that autonomous control systems are optimized for lunar microgrid operations and contribute to the overall success of future lunar missions.

As humanity prepares for long-term lunar exploration, sustainable energy solutions are crucial for supporting lunar habitats, life support systems, and scientific research. Direct Current (DC) microgrids are emerging as an ideal solution due to their efficiency and compatibility with solar power systems. However, the harsh lunar environment—characterized by extreme temperatures, high radiation, and lunar dust—presents significant challenges for DC grid protection. This paper explores the need for robust protection mechanisms in lunar microgrids, reviewing existing terrestrial DC protection strategies, their limitations, and necessary adaptations for lunar applications. It examines key challenges such as lunar dust accumulation, radiation exposure, and micrometeorite impacts, proposing solutions for each. Additionally, the paper discusses the potential of AI and machine learning to enhance fault detection and system optimization. A case study is presented that explores the adaptation of Earth-based protection technologies to lunar microgrids, highlighting performance evaluations and identifying key areas for future research. The goal is to develop resilient, efficient, and autonomous DC protection systems that ensure reliable energy supply for future lunar missions.