5G networks provide the Quality of Service (QoS) to various services, including Enhanced Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communications (URLLC), and massive Machine-Type Communications (mMTC) with different requirements. However, the existing 5G network deploys the static 5G QoS Identifiers (5QI) framework, which provides a basic level of services differentiation and supports the class-based prioritisation. The static 5QI fails to support the dynamic and user-centric QoS requirements. This paper reviews user-centric QoS provisioning schemes that focus on the Software-Defined Networks (SDN), Machine Learning (ML), and network slicing schemes. A structured evaluation methodology is used to analyse the existing works in terms of architecture, mechanisms, QoS metrics, scalability, and real-world feasibility. The reviews discover that SDN improves programmability and policy-based control, ML predicts the traffic adaptively, and network slicing provides service isolation. However, user-centric schemes can increase computational load, raising controller overhead and latency. This paper identifies the research gaps in the user-centric schemes related to scheme readiness, scalability, security, and heterogeneity. The paper also proposed a direction for deploying user-centric QoS schemes in real 5G networks that support future 5G enhancement that covers the adaptive, efficient, and personalised QoS provisioning.
Soil infiltration plays a crucial role in the hydrological cycle, water resource management, and sustainable agricultural practices. It is important to identify the emerging trends and research gaps for this study. Therefore, this study aims to present a comprehensive and systematic academic review of soil infiltration in the field using bibliometric analysis with the aid of VOSviewer to evaluate global research trends in soil infiltration based on author keywords, affiliated countries, and co-authorships. This study retrieved 494 articles published between 2000 and 2025, as of January 16th, 2025, from the Scopus database. Results have shown a steady decrease and increased for annual number of publications, with the lowest drop in 2024 with 22 publications and the highest rise in 2021 with 30 publications. The leading contributors are from China and the USA, leading by 28% and 21% respectively. Meanwhile, Nongye Gongcheng Xuebao/Transactions of the Chinese Society of Agricultural Engineering contributed the most publications. Although the Science of The Total Environment (STOTEN) journal ranked second after Water (Switzerland) in the top journal, the CiteScore 2023 for STOTEN is the highest with 17.6. This study underscores the effectiveness of a water management system, which includes it supports for flood risk mitigation, enhances agricultural productivity and improves groundwater recharge. Future research should focus on optimising the soil infiltration model by adapting AI tools to consider climate change factors
Microplastics (MP) have actively polluted various water sources, including oceans, rivers and lakes. MPs are tiny plastic particles, typically less than 5 millimetres in size. They originate from degraded larger plastics or microbeads in products. It can enter human bodies through ingestion or inhalation. Their small size allows them to penetrate tissues and organs, potentially causing inflammation. In the environment, MP accumulates in soil, water, and air. They can disrupt ecosystems by entering the food chain, affecting aquatic life. The objective of this paper is to provide a comprehensive review of established MP identification methods and to recommend new technologies for future environmental monitoring. The methodology employed a comprehensive literature review and comparative analysis of current MP detection techniques. The non-optical analytical tools, such as Pyrolysis-Gas Chromatography-Mass Spectrometry (Py-GC-MS), Nuclear Magnetic Resonance (NMR), X-ray Diffraction (XRD), Thermogravimetric Analysis (TGA), and Scanning Electron Microscopy (SEM), alongside prominent optical methods, include standard Fourier Transform Infrared (FTIR), Attenuated Total Reflection Fourier-Transform Infrared Spectroscopy (ATR-FTIR), Raman spectroscopy, Nile Red fluorescence, and UV-Visible (UV-Vis) spectroscopy. The results obtained conclude that no single method is perfect, but Micro-Fourier Transform Infrared (μ-FTIR) spectroscopy is highly recommended as the primary optical sensing approach for standardised monitoring due to its optimal balance of speed, reliability, and accessibility. The study also highlights that the integration of Artificial Intelligence (AI) and Machine Learning (ML) is the future of optical MP detection, capable of automating complex spectral matching and noise reduction. Potential applications of the research proposed multi-modal framework can be directly applied to large-scale, cost-effective environmental monitoring, enabling broader geographical coverage and aiding policymaking, particularly in developing nations.
A booming population around the world raises the concern of shortages of food resources in this new era. Thus, monitoring and managing crop production is extremely essential, especially rice crops, as they are the fundamental food source for most countries. Several challenges need to be addressed in this case, such as the classification of farmland from various land usages, precise monitoring of rice seedlings, and segmentation of rice growth. By leveraging advanced technologies such as drone imagery and machine learning, this paper proposed a new integrated pipeline for rice field classification and growth monitoring: a combination of convolutional neural networks (CNNs), You Only Look Once (YOLO), and modified U-Net models. These models were used in stages, specifically for paddy field classification, rice seedling detection, and rice growth segmentation. Substantial measurements and analysis have been carried out to verify the performance of the proposed system, including an accuracy of at least 85%, low classification/segmentation loss below 0.35, and high detection recall above 0.9. Thus, the findings highlight how combining different machine learning models with aerial photography can revolutionise conventional farming methods for better efficacy.
Vehicle number is an essential issue for intelligent traffic monitoring systems. It plays a significant role in the transportation sector. In addition, recent population growth and human needs have increased due to the use of vehicles. Therefore, controlling traffic is becoming a complex problem to solve. An Automated Plate Number Recognition System (ANPR) program is needed for traffic control. This research includes developing and implementing a multi-stage preprocessing methodology with various techniques to identify vehicles’ plate numbers. The detection system can achieve successful vehicle management: car parking as a case study. The proposed ANPR system employs a template matching algorithm based on a multi-stage preprocessing technique for character identification. Those techniques include noise filtering, component isolation, and correlation-based similarity measurements. The methodology process steps are grayscale conversion, binarisation, bilateral filtering, edge detection, contour localisation, segmentation using connected components, and character recognition through cross-correlation. The system was evaluated on 20 Iraqi one-line plates and achieved a recognition accuracy of 91%.
Integrating serious games in educational settings has emerged as a promising approach to enhance learning experiences. In serious games, gesture-based interactions, such as motion sensors and wearable devices, enhance intuitiveness and engagement. The lack of standardisation in educational user engagement evaluation methods for serious games and gesture-based interaction has been highlighted in multiple studies. Most standardisation splits the dimensions, complicating the evaluation process. A systematic mapping study was conducted to address this issue, examining serious games and gesture-based interaction in education from 2018 to 2024 across five electronic databases. Despite wide interest among authors, there has been a limited systematic effort to define the concepts. This paper contributes by identifying elements used in serious games and gesture-based interaction for education, analysing research approaches and contributions, and proposing design criteria for evaluating educational prototypes based on user engagement. The study addresses differing opinions among authors and suggests solutions to unify concepts related to user engagement. The paper recommends several guidelines for serious games and gesture-based interaction in education, along with instruments to enhance user engagement.
This review explores the key components of crop production and energy consumption in indoor farming, where it usually consists of systems such as LED lighting, ventilation (fans), humidity and hydroponics systems to offer controlled environments for crop growth. However, these systems consume large amounts of energy. Thus, it is necessary to reduce this energy consumption in crop production. This review examines three modelling approaches for energy optimisation, namely: white box (physics-based), black box (data-driven), and grey box (a hybrid of the two). Each method's strengths and limitations are discussed in terms of their application to indoor farming. The findings emphasise the importance of selecting the right optimisation model based on specific goals and resources, with the potential to significantly improve energy efficiency in indoor farming. Further research is encouraged to refine these models and support sustainable food production.
Machining is essential in manufacturing for shaping materials by removing excess through contact between the tool and workpiece. However, dry machining generates significant heat and cutting forces, leading to tool degradation and surface imperfections. While flood cooling helps address these issues, it presents environmental and health concerns due to large volumes of fluid and harmful chemical additives. Minimum Quantity Lubrication (MQL) offers a greener alternative by applying only the minimum amount of lubricant directly to the cutting area. Still, it frequently proves insufficient in handling difficult-to-cut materials like nickel-based alloys, owing to poor heat removal and lubrication. This study investigates enhancing MQL using silicon dioxide (SiO2) nanoparticles in vegetable oil-based formulations, creating an improved nanolubricant system (NMQL). The volume concentrations of five different percentages were tested: 0.01%, 0.03%, 0.05%, 0.07%, and 0.1 % of SiO2 biodegradable nanolubricant. Stability was evaluated through visual sedimentation observation, UV-visible measurements, and zeta potential analysis. Some slight sedimentation occurrences through visual observations indicated nanolubricant stability was measured at a 0.1% volume concentration, showing a linear correlation adhering to Beer-Lambert law principles. The most stable nanolubricant resulted from 2 hours of sonication at a 349 nm wavelength, exhibiting an excellent zeta potential stability of 98.3 mV. Thermophysical testing revealed that 0.1% concentration produced the highest dynamic viscosity compared to the base oil. In the tribological evaluation, the 0.07% concentration exhibited the lowest coefficient of friction and wear scar diameter. Overall, biodegradable nanolubricants serve as a potential for Minimum Quantity Lubrication (MQL) machining.
In view of the fundamental limitations of standard machine learning methods, which primarily depend on actual and labelled attack data, the identification of novel and zero-day cyber-attacks continues to be an essential and continuing difficulty in cybersecurity. Techniques become vulnerable to emerging attack vectors, considering these models often do not apply beyond identified threat classes. To overcome this limitation, present ZTFER (Zero-Shot Threat Identification through Foundation model-aided Embedding and Reasoning), a Zero-Shot threat identification model which utilises the basis model semantic embedding to detect threats which had not been observed before without needing retraining. Through combining natural language threat characterisations and real-time system activity into a common semantic space, ZTFER implements zero-shot learning and makes it achievable to classify unnoticed threats by considering contextual similarities. Furthermore, present A-SENT, a real-time responsive inference method which dynamically evaluates and addresses threat conduct through integrating Large Language Model (LLM) reasoning in real-time threat analysis updates. The proposed ZTFER model obtained a zero-day detection score of 58.9% and an accuracy of classifying 91.3%, outperforming the traditional and few-shot baselines. The experimental results demonstrate that ZTFER can be more effective and have better generalisation capability for detecting known and unknown cyber-attacks. The proposed framework does away with the requirement for continuing retraining and enables quick response to emerging threats. Creating powerful, scalable, and smart defence mechanisms able to recognise and understand new security threats in real time becomes achievable in a significant way by this research.
Early identification of durian cultivars at the seedling stage remains challenging because conventional methods rely on fruit characteristics that appear only after several years, often leading to misdistribution and inefficiencies in certification. This study proposes a real-time, non-destructive deep learning framework for intra-species durian cultivar identification using leaf images. A dataset of 1,788 leaf images from five cultivars (Bawor, Kani, Monthong, Musang King, and Petruk) was collected under real-world conditions. Unlike existing studies that formulate plant varietal identification solely as an image classification task, this work introduces a detection-based formulation that explicitly models cultivar-specific leaf features at the object level. This formulation enables the model to localise and learn fine-grained morphological differences that are often overlooked in global classification approaches. To validate this approach, MobileNetV2, Xception, and YOLOv11 were systematically evaluated using K-Fold Cross Validation under the same experimental setting. The key contributions of this study are: (1) the formulation of intra-species plant cultivar identification as an object detection problem rather than a pure classification task, (2) the development of a practical, real-time framework for early-stage durian cultivar identification using leaf images in unconstrained environments, and (3) a comparative analysis demonstrating the effectiveness of detection-based models in capturing subtle inter-varietal differences. Experimental results show that YOLOv11 achieves over 99% validation accuracy, with precision and recall above 0.95, while maintaining an inference time of 0.2 ms per image. These findings highlight the potential of detection-based approaches for AI-assisted seedling selection in horticulture.
Intuitive visualisation tools are essential for the effective coordination of construction projects to convey complex spatial and technical information to diverse stakeholders. Traditional design documentation using 2D drawings and static 3D models cannot accurately depict complex spatial relationships and construction sequences. This results in design misinterpretations, inefficient constructability reviews and rework costs. Most of the current extended reality (XR) deployments in construction are essentially used as visualisation tools. Gamification usage in the industry is mainly oriented to the educational workflow context. This study develops an Extended Reality Information Space (XRIS), a HoloLens-based system that spatially anchors construction data onto physical environments. The system operates through gaze-triggered information tags and embedded game-based mechanics. Three functional layers covering immersive visualisation, information management, and interaction and gamification were comprised in this system. Semi-structured interviews were conducted with eight construction professionals and two XR specialists to identify the benefits of XRIS implementation and the impact of gamification on XRIS. Data were then analysed using thematic analysis. The findings suggest that XRIS enhanced spatial understanding, user accessibility, team communication, and motivation. Gamification integrated in the visualisation layer maintained the users’ engagement through the behaviour of spatial exploration and task completion. Both mechanisms supported deeper retention and more precise articulation of design logic than passive documentation review. This study contributes a professional-oriented, empirically evaluated XRIS configuration advancing immersive tool design for construction practice.
Large language models (LLMs) such as BERT (Bidirectional Encoder Representations from Transformers) have caused strides of progress in the field of natural language processing (NLP), with lightweight variants such as DistilBERT, MobileBERT and TinyBERT being developed to lower the resource requirements to deploy the models in real-world settings. However, while past research has investigated improving non-distilled models on a variety of benchmarks by changing their architecture, there are limited studies that explore how the lightweight variants may perform in the same conditions. This study applies modifications and ensemble techniques on lightweight BERT models in extractive question answering (QA) to address this gap. The experiments were conducted using three datasets: SQuAD, AdversarialQA, and a newly curated Sexual and Reproductive Health QA (SRHQA) dataset consisting of 1000 samples. From the results, it was shown that applying the same modifications that would enhance the base BERT models to the lightweight variants generally caused a 0.14 - 5.20% F1 decrease in performance, with marginal exceptions observed in specific dataset-model combinations. Ensembling, on the other hand, showed improvements across all the datasets, ranging from 2.19 - 19.46% F1 over the BERT baseline. The results of the study highlight the sensitivity of the lightweight models, their trade-offs with efficiency, and that an ensemble is a valid approach to utilising the lightweight models without architectural modifications.
Speech communication involves the exchange of information between two or more individuals; however, background noise and reverberation often degrade speech clarity and intelligibility. The impact of these degradations depends on factors such as the number, intensity, and spatial characteristics of noise sources, as well as reflections from surrounding surfaces. These effects pose significant challenges for applications including teleconferencing, hearing-aid devices, and human-machine voice interfaces. This study aims to address the combined effects of background noise and reverberation by proposing a speech enhancement framework that integrates Minimum Variance Distortionless Response (MVDR) beamforming with a Coherent-to-Diffuse Power Ratio (CDR) based post-filter. The methodology relies on parallel processing, where MVDR beamforming performs spatial noise suppression, while CDR values are estimated from microphone-domain signals and used to compute post-filter gains for suppressing residual diffuse noise and reverberation in the beamformer output. The proposed algorithm was evaluated over an input signal-to-noise ratio range from 0 dB to 40 dB and compared against the Integrated Sidelobe Cancellation Linear Prediction (ISCLP) baseline using four objective metrics: Perceptual Evaluation of Speech Quality (PESQ), Extended Short-Time Objective Intelligibility (ESTOI), Cepstral Distance (CD), and Weighted Spectral Slope Distance (WSS). The results demonstrate consistent improvements in PESQ and reductions in CD and WSS across all tested conditions, while gains in ESTOI remain modest. These findings indicate improved spectral fidelity and speech naturalness, highlighting the practical relevance of the proposed framework for real-time, low-complexity speech enhancement in noise and reverberation-prone communication systems.
Blockchain is an efficient method to manage and secure data, but scalability remains a limitation. The proposed work concentrates on a novel consensus algorithm, Proof of Useful Work-Authorisation-Storage Availability. The work emphasises the scalability issues and trust-related issues in blockchain-based smart contracts. The primary step in the proposed technique is to verify authorisation by the hash code of the preceding block to generate a digital signature. After information is legitimised, check whether enough storage space is available or not. Transactions are recorded to a block only after they are validated and there is enough space. A transaction is included in a block only after it is mined. To compare how efficient the proposed consensus is, in terms of energy consumption, delay, and the number of transactions it can process, the study uses Python and Solidity. The new method reduced computational energy consumption by 39% and increased transaction capacity by 17%, and decreased network delay by 49% compared to the proof of useful work algorithm. This is an indication that the efficient consensus will help us scale without compromising security and decentralisation. The contribution is beneficial in determining which nodes to apply in the consensus layer. The system can help with banking applications and supply chain systems that are blockchain-based. It highlights three key points: scalability, low delay, and efficient consensus, which are necessary for these systems to function properly.
High-Speed Rail (HSR) projects involve complex systems, huge investments and multidimensional risks. Effective risk reduction requires context-sensitive mitigation strategies which are accurate. The traditional entropy-based models are static and based on the subjective judgment of humans, and they lack adaptability in real-time modelling of the complexity of the HSR environment. The paper proposes an innovative integrated entropy-based risk assessment and mitigative framework, which is related specifically to HSR systems and will centre around dynamic gradations of risk weights and mitigation strategy. The framework presented is a series of 5 interconnected methodology blocks. First Dynamic Spatio-Temporal Entropy Weighting (DSTEW) generates its entropy weights, which are real-time adaptable timeline or temporal, e.g. seasonal, operational time-line factors, and spatial, i.e. zone-related variability. For checking the robustness of the procedure, Multi-Entropy Cross Validation (MECV) is used to check the consistency of subjective entropy estimates relating to the various spatial zones and time zones to retain only statistically consistent weight vectors. Entropy-Driven Bayesian Risk Adjustment (EDBRA) then adjusts and modifies these weights by including the history of past risk occurrences by means of Bayesian updating of risk. Subjective uncertainties are also resolved by using Hybrid Fuzzy-Entropy Multi-Criteria Prioritisation (HFEMCP), which uses the system of fuzzy logic in its consideration of the recalibrated weights and thus generates risk matrices of prioritised risks. Lastly, Entropy Resilient Networks Modelling (ERNM) produces inter-risk influence networks in which mitigation strategies are determined in a global way based on the various centralities which are derived from the entropy weighting procedures. The simulation results feature a quantum leap in improvements that include 10% to 15% improvements in weighting accuracy, validation consistency >95% and systemic resilience improvements of 20% to 30%. Thereby allowing for an adaptive, evidence-based, system-wide risk reduction planning of risk in HSR projects.
Damage to the bearings of a ship's propulsion motor represents a significant reason why ships experience operational disruptions. Therefore, it is important to detect and diagnose early on problems related to the bearings of a propulsion motor early on to reduce maintenance costs and prevent potential maritime accidents. Additionally, since bearings subjected to variable loading and speed in dynamic environments have the possibility for various types of damage to occur, an accurate, non-destructive detection method is required to allow for early indication of damage to bearings. A condition monitoring system based on acoustic signals was developed within this study to be able to monitor the health of the bearings of a propulsion motor without damaging them. A method called Short Time Fourier Transform (STFT) was used to analyse the time vs frequency characteristics of the acoustic signal generated by the motor during various operating conditions. The STFT analysis revealed unique spectral patterns that identified different forms of bearing damage, i.e., failure of the inner-race, outer-race, ball of the bearing, and cage. Testing of the system was conducted with a variety of combinations of speed and load to identify the reaction of the system to variations in operational conditions. Monitoring ship propulsion motors using the STFT algorithm has been proven to be suitable for processing non-linear data with a high accuracy of 97.5%. Development of this system will provide a safe and environmentally friendly way to monitor the operation of equipment at sea.
Plant-derived lipophilic bioactive compounds often show poor dispersion and instability in a water system; this limits their use in topical and pharmaceutical formulations. This study aimed to develop and optimise an oil-in-water nanoemulsion containing Strobilanthes crispus extract dissolved in red palm oil using Response Surface Methodology (RSM). A three-factor Box-Behnken design was applied to evaluate the effects of Tween (R) 80 concentration, glycerol concentration, and homogenisation pressure on droplet size, polydispersity index (PDI), and creaming index. The results show that the surfactant concentration was the most important factor affecting the droplet size and overall stability. The optimised nanoemulsion formulation comprised 5.06% (w/w) Tween (R) 80, 10.4% (w/w) glycerol, and a homogenisation pressure of 617.1 bar. Experimental validation produced a nanoemulsion with a droplet size of 151.8 +/- 1.3 nm, PDI of 0.127 +/- 0.021, and no visible creaming. The regression models showed strong predictive ability for droplet size (R2 = 0.984) and creaming index (R2 = 0.999), but moderate prediction for PDI (R2 = 0.7516). Stability studies under different pH (2-6), temperature (40-60 degrees C), ionic strength (0.5-2.5mM), and centrifugal conditions (1000-5000rpm), as well as four-month storage, showed that the nanoemulsion remained physically stable without phase separation. This stability is mainly attributed to steric stabilisation from the non-ionic surfactant. Overall, the study demonstrates that Response Surface Methodology is an effective tool for optimising nanoemulsion formulations for the delivery of lipophilic plant bioactives.
The severe stage of diabetic retinopathy (DR) is often identified by the presence of hard exudates on ophthalmologic examination. For precisely detecting exudates, Optic Disc (OD) segmentation is significant due to the high similarity between exudates and the OD. The primary objective of this OD-SAM study is to examine the feasibility of a zero-shot framework for OD segmentation, intending to improve subsequent hard exudate detection in retinopathy screening. The proposed work integrates automatic OD localisation using the peak-end thresholding approach, prompt-based optic disc segmentation using the segment anything model (SAM), a multi-criteria decision-making (MCDM) approach for optimal OD mask selection, and ellipse fitting to smooth the disc boundary. Subsequently, the OD is removed to evaluate its impact on YOLO-based hard exudate detection. The efficiency of this proposed framework is tested on the IDRiD dataset. It achieved an 86.2% overlap and a 90.7% dice coefficient. The results show the effectiveness of the proposed approach for OD segmentation and highlight improvements in exudate detection after OD removal. This study can support lesion analysis and grading in an automated diabetic retinopathy screening and severity assessment.
Autism Spectrum Disorder (ASD) is a complex, long-lasting neurodevelopmental condition characterised by communication difficulties. However, early identification of ASD may aid in the design of appropriate treatments to improve communicative development. This study employs association rule mining algorithms, such as apriori, predictive apriori, and Tertius, to identify age-specific behavioural markers across four developmental stages: toddlers, children, adolescents, and adults. These algorithms identified several key behavioural indicators associated with ASD, including pretending games (A5), activity switching (A4), difficulty in making friends (A10), difficulty in conversations (A5), detection of listener boredom (A6) and easy reading of emotions (A9). The apriori algorithm achieved the highest confidence of 100.00% for the adolescent dataset, whereas the predictive apriori algorithm demonstrated the highest accuracy of 99.50% for the toddler dataset, 99.40% for the child dataset and 99.39% accuracy for adult dataset. In contrast, the Tertius algorithm showed the highest confirmation rate of 66.8% for the child dataset, although it required the most time for manipulation across all datasets. Our findings demonstrate that rule mining effectively uncovers clinically relevant patterns, with behavioural questions proving more significant than demographic factors, such as gender or birth complications. Furthermore, no connection was observed between ASD and a history of jaundice at birth. Performance comparisons among the three algorithms are also presented. The rules generated through our investigation will help physicians in the early detection of ASD, thus paving the way for a timely and targeted intervention.
Large campus’s navigation becomes a crucial issue nowadays due to its layout complexity and digital services’ diversities. Addressing this challenge, this research focusses on Universiti Teknologi Malaysia and presents a novel campus navigation solution that incorporates location-based Augmented Reality with an Artificial Intelligence chatbot. In contrast to previous AR navigation systems, which rely on static visual markers, and chatbot systems, which provide text-based support, the proposed system links real-time AR guidance with intelligent and context-sensitive handling of inquiries. This dual functionality capability sets the solution apart because users can follow AR-driven directions, ask questions, and receive instant information about campus facilities. The prototype was developed using an Agile-inspired iterative methodology that allows for continuous refinement based on user feedback. Evaluation results show a high degree of user satisfaction: 95% of respondents agreed that the system enhances the clarity of wayfinding, while 70% reported that they feel more confident in navigating a campus independently. The study emphasises the practical impact of using combined AR and AI technologies for campus digitisation by showing that the system can be effective in improving navigation efficiency, reducing reliance on physical signage and human assistance, and making the process more engaging and interactive for new students and visitors. In this work, the authors provide a novel model of a hybrid AR-chatbot navigation that can be used in future initiatives on smart campus development and scalable digital support systems in higher education.
