Electric vehicles (EVs) significantly improve environmental sustainability by eliminating exhaust emissions, therefore enhancing air quality and decreasing greenhouse gas emissions. The efficacy and performance of electric vehicles are largely contingent upon their electric motors. Induction motors (IMs) have significant advantages for EVs, such as control flexibility, cost-effectiveness, and improved thermal management, rendering them a favoured option for traction applications. This study presents a 5 Horsepower (HP) three-phase induction motor employing novel Carbon Nanotube-Copper (CNT-Cu) as alternatives to traditional conductive materials, such as copper, in electric vehicles. CNT-Cu improves conductivity, thermal stability, and weight reduction of IM windings. The design phase utilises the ANSYS RMxprt tool, succeeded by a comprehensive finite element analysis (FEA) with ANSYS Maxwell 2D, guaranteeing precise simulation-based validation of the proposed motor design. The simulated results include efficiency, output power, torque, and air gap flux density, indicating that CNT-Cu composites enhance electrical and mechanical performance relative to traditional copper windings. The findings have been compared with traditional copper-wound induction motors, confirming the viability of CNT-Cu composites for high-performance electric vehicle motor applications. The results validate that CNT-Cu composites have considerable promise for advanced electric motors, enhancing energy efficiency and reducing weight in electric vehicle powertrains.

Non-technical losses (NTL) in power distribution, such as illegal meter tapping, cause significant financial losses for utilities, amounting to billions annually. This study evaluates various machine learning methods for NTL detection, addressing the challenge of imbalanced electricity consumption data. Seven techniques for data balancing were employed: Adaptive Synthetic Sampling (ADASYN), Random Over Sampling, Random Under Sampling, Near Miss Under Sampling, and several variations of Synthetic Minority Over Sampling (SMOTE), including Borderline-SMOTE, SMOTE-ENN, and SMOTE-Tomek links. The model comprises two stages: first, seven classification algorithms (Decision Tree, Logistic Regression, XGBoost, Random Forest, SVM, Naïve Bayes, and KNN) were tested across diverse training-testing ratios to identify optimal performance. The second stage applied the comprehensive consumption dataset along with data balancing techniques to improve algorithm efficacy. Performance metrics—accuracy, precision, recall, F1 score, and Matthews Correlation Coefficient (MCC)—were utilized for evaluation. Results revealed that the Random Forest algorithm, when paired with Random Over Sampling at a 70 − 30% training-testing ratio, yielded the highest metrics: 98.03% accuracy, 99.02% precision, surpassing existing literature. The model achieved exceptional precision (0.990) and the highest overall performance, with rigorous statistical testing confirming all improvements were significant at the 95% confidence level.

Precision and smoothness in controlling mechanism is a mandatory requirement for those robotic applications, which are associated with upper limb rehabilitation and comprise of five degree of freedom. This study comprises of analysis of a vast range of techniques of optimization, for determining ideal parameters of a Proportional Integral Derivative (PID) controller, which is managing a dynamic model of the upper limb rehabilitation system. The control technique which has been proposed was carried out on a five-degree of freedom (DOF) hardware, and priory to any trials under clinics, more than five healthy subjects underwent trials following strict protocols. There exist evidences of rejection to the external disturbances by the control technique, alongside keeping the system stable during the course of harsh and extremely uneven circumstances. This research study validates the mathematical models for finding optimizing parameters of the PID, including a comparative study comprising of four diverse cost functions. The objective functions include Integral Absolute Error (IAE) and Integral Squared Error (ISE). A variety of PID tuning methods were assessed based on core performance indicators such as Integral Square Error (ISE), Integral Absolute Error (IAE), overshoot, rise time, and settling time. The PID controller tuned using Particle Swarm Optimization (PSO) struck a good balance, with relatively low IAE, moderate overshoot (4.51%), and a decent settling time of 8.28 s. The Firefly algorithm also delivered promising results, achieving the shortest rise time (0.689 s) and a quick settling time (1.08 s), while keeping the overshoot (4.45%) nearly on par with PSO. In comparison, Ant Colony Optimization (ACO) produced a significantly higher overshoot of 29.6% and longer settling behavior. The Artificial Bee Colony (ABC) method stood out with the lowest settling time (0.84 s) and minimal overshoot (4.29%). The classical Ziegler-Nichols approach, however, showed poor performance, with a high overshoot of 55.3% and slower system response. Overall, these outcomes suggest that nature-inspired techniques, especially Firefly and ABC, can offer more efficient and stable control than conventional tuning methods.

The emergence of sixth-generation (6G) wireless networks introduces unprecedented requirements for ultra-secure, low-latency communication across heterogeneous space–air–ground integrated network (SAGIN). Existing drone communication frameworks including LoRaWAN, Long Term Evolution, and Ad Hoc mesh architectures exhibit critical vulnerabilities to eavesdropping, jamming, and quantum-computational attacks due to their reliance on classical cryptographic primitives. To address these challenges, this work presents the Quantum-Secured Adaptive Routing Algorithm (QSARA), a novel framework designed for 6G-enabled unmanned aerial vehicle (UAV) networks that integrates Quantum Key Distribution (QKD), Reconfigurable Intelligent Surfaces (RIS), and Joint Communication and Sensing (JCAS) to enhance information-theoretic security and real-time performance. The proposed framework employs a quantum-augmented dynamic graph model to represent UAV swarm networking and uses Proximal Policy Optimisation (PPO)-based deep reinforcement learning to optimise routing under adversarial and uncertain conditions. A multi-objective cost function jointly captures classical quality of service metrics, such as latency, bandwidth, and energy consumption alongside with quantum-layer security indicators, including quantum bit error rate, key pool entropy, and key availability. High-fidelity simulations with 500 mobile drones under diverse adversarial threats demonstrate that the proposed framework achieves a key establishment success rate of 96.2%, end-to-end latency of 23.7 milliseconds, energy consumption of 7.8 watt-hours, and a packet delivery ratio of 94.1%, outperforming state-of-the-art classical and quantum-aware baselines. These results position the QSARA as a scalable and quantum-resilient routing solution for mission-critical UAV networking in next-generation 6G smart mobility ecosystems.

Camouflaged object detection particularly is considered as challenging and crucial because these objects are designed to either mimic their environment or be completely hidden within it. The goal of camouflage patterns utilization is to help objects blend into their surroundings, making them harder to detect. One of the biggest hurdles is distinguishing the object from the background. Many efforts have been made to tackle this problem all around the globe, and this research builds on those advancements. The focus is on developing methods for detecting camouflaged targets in military settings, including materials, operations, and personnel using convolutional neural network. A key contribution of this work is the MSC1K dataset, which includes 1,000 images of camouflaged people with detailed annotations for object-level and bounding-box segmentation. This dataset can also support broader computer vision tasks like detection, classification, and segmentation. Additionally, this research introduces the Dynamic Feature Aggregation Network (DFAN), a method inspired by previous studies that uses multi-level feature fusion to detect camouflaged soldiers in various conditions. Extensive testing shows that DFAN and SINet-V2 (Search and identification network) achieved the highest accuracy with the least error, while SINet struggles the most. Notably, DFAN shines with its precision-recall balance, while SINET lags behind, potentially due to difficulties in handling intricate saliency patterns. The most intriguing contrast arises in the third setting (MSC1K + CPD), where DFAN remarkably excels, displaying superior structural similarity, strong human-perception alignment, and optimal precision-recall trade-offs. DFAN emerges as the top performer in terms of error minimization, achieving the lowest MAE values: 0.051 for MSC1K, 0.004 for CPD, and 0.028 for the combined dataset. In contrast, SINet shows the highest error rates, making it the Least reliable model, with MAE values of 0.079, 0.157, and 0.049 respectively. ZoomNet and SINetV2 delivered moderate performance; ZoomNet records MAEs of 0.056, 0.005, and 0.029, whereas SINetV2 reports 0.051, 0.005, and 0.027 in the same settings. These results indicated that DFAN and SINetV2 consistently produced more accurate predictions, while SINet has less precision. Overall, the comparative assessment sheds light on how each model adapts to varying datasets, revealing key insights into their performance robustness.

Protein function categorization or prediction is considered to be one of the key subfields of bioinformatics and computational biology. Clarifying the relationships and functions of the proteins that essentially comprise the body is beneficial. Proteins, one of the biggest types of macromolecules, are essential to every biological function. They consist of amino acids, which are believed to be the main components of proteins. Every protein has a distinct amino acid sequence that dictates its structure and function since all proteins are made of amino acids, just as all nucleic acids are made of nucleotides. There are several ways to classify proteins, such as grouping them into several groups according to their sequences, physiological structures, and functions. Recent advances in machine learning and deep learning models have made it possible to classify proteins in a number of ways. In light of this, the primary objective of the current work was to assess how well various ML and DL algorithms performed on the Protein family database (PFAM), which serves as a reference dataset for the protein classification problem. Numerous sequences connect each protein to one of the numerous predetermined groups. Consequently, the study's findings demonstrated significant performance disparities between deep learning models like 1D Dilated CNN and machine learning models like SVM, Random Forest, and Naive Bayes. The best tested architecture was the 1D Dilated CNN, which achieved the maximum accuracy of 97%. It is among the most scalable models available, needing the least amount of time and 6%. As demonstrated in the instance of protein classification, deep learning is actually best applied in scenarios where it has the highest chance of success, such as in datasets with a high dimensionality and a large number of features.