Ibrahim A. Hameed

Accurate forecasting of daily solar radiation is essential for renewable energy management, optimization of photovoltaic systems, and power-grid stability. This study presents an evaluation of six forecasting models: SARIMA, LSTM, NARX, Random Forest, XGBoost, and a stacking ensemble. Daily solar radiation and meteorological data were retrieved from the NASA Prediction of Worldwide Energy Resources (NASA POWER) platform for Poruba, Ostrava, Czech Republic. A chronological data-partitioning strategy was adopted, with data from 2001–2020 used for training, 2023 for validation, and 2024–2025 reserved for prospective testing. The modeling framework incorporated meteorological variables, cyclic calendar encodings, lagged radiation observations, and rolling statistical features to capture short-term fluctuations and long-term seasonal dynamics. Among the individual models, XGBoost achieved the strongest test performance (R² = 0.7704), followed by Random Forest (R² = 0.7565), NARX (R² = 0.7524), and LSTM. SARIMA exhibited the weakest predictive accuracy (R² = 0.7265), reflecting its limitations in modeling nonlinear temporal dependencies. The stacking ensemble, combining Linear Regression, Random Forest, and K-Nearest Neighbors as base learners and Ridge regression as the meta-learner, achieved the highest performance (R² = 0.9997), outperforming all individual models on the prospective test set. Statistical significance testing using the Friedman omnibus test, Wilcoxon signed-rank test, and Diebold–Mariano test with Holm–Bonferroni correction confirmed that the ensemble’s superiority was statistically robust across all pairwise comparisons (p < 0.001). These findings demonstrate that ensemble learning, supported by structured feature engineering and temporally consistent model evaluation, provides an effective and generalizable framework for long-term solar radiation forecasting.

Background The global shift towards an ageing population and workforce shortages increases demands on health and social care systems. Proactive, personalized, and resource-efficient care models are therefore needed. Continuous monitoring using wearables, together with artificial intelligence (AI), offers potential for early detection of health deterioration and timely intervention. Still, existing solutions often target single diseases and lack integration into clinical workflows. Objective This article presents the study protocol for the project Predicting Care Needs of Older Adults in the Healthcare System through AI-Enabled Analysis of Patient Monitoring Data (PAI), which aims to develop an AI-driven predictive model based on data from wearables (smartwatches) to identify health decline among older persons in the near future. In addition, the project will evaluate the pilot implementation of this model within health and social care settings in three countries and generate recommendations for future adoption. Methods To answer the project aim, the AI model will be developed using historical datasets from wearables linked to emergency health care visits. To guide and evaluate the pilot implementation in Sweden, Norway, and Spain, staff from health and social care, managers, and IT professionals will be interviewed before implementation to explore current work practices and relevant contextual factors. Users of the smartwatch (older persons) will also be interviewed to explore their experiences, needs, and expectations. Follow-up interviews will be conducted after implementation to capture users’ and health and social care staff’s experiences of use. In addition, quantitative data on the AI model’s performance will be collected. The recommendations will be developed in an iterative process, including both researchers and stakeholders. Expected Results The project will identify key data requirements, deliver a predictive AI model, and develop evidence-based recommendations to support the responsible and scalable implementation of the proposed AI-solution in healthcare and social care settings. Conclusions By combining technical innovation with real-world implementation research, the PAI project seeks to advance the use of AI in care for older persons, support proactive and person-centered interventions, and contribute to more sustainable healthcare systems.

Non-contact laser triangulation is widely used for high-precision measurement and inspection across diverse industrial applications. Although these systems offer speed, accuracy, and non-invasiveness, their reliability is highly sensitive to multiple influencing factors. These include surface characteristics, material transparency, environmental conditions, and measurement configurations such as roughness, illumination, and incidence angle. This study systematically examines the impact of these factors on measurement reliability through repeated experiments conducted in a highly configurable test environment. The results identify key sources of variability and provide practical recommendations for selecting suitable measurement configurations under different operational conditions. By applying these insights, engineers and practitioners can optimize laser triangulation systems for enhanced reliability in demanding real-world applications.

This paper investigates the latest advancements and industrial applications of Digital Twin (DT) technology, focusing on 3D behavior representation across the entire lifecycle, from design and initial planning to testing, implementation, and optimization. In this comprehensive review, the authors define 3D behavior as the virtualization and simulation of controlled technologies, encompassing processes performed, movement scenarios of the various parts of the technology, and their responses to the actual states and the surrounding environment. The study also explores the industrial applicability of 3D modeling and simulation using the latest software tools and methodologies. Furthermore, the authors define several use cases, derived from their own research experiences, that illustrate the different variants of Digital Twins across various stages of development and production within digitalized technology. These use cases highlight how Digital Twin solutions are applied in real-world scenarios, aligning with current digitalization trends and Industry 4.0 standards. The findings are discussed from an innovative perspective, emphasizing how Digital Twin technology can significantly improve all phases of industrial digitalization. This paper not only examines the research, theoretical foundation, and historical development of industrial Digital Twins but also clarifies key concepts such as modeling and model-based design, while providing an overview of relevant software tools. In addition, five detailed industrial use cases are presented, illustrating the practical implementation and benefits of Digital Twin technology in various industrial applications.

Accurate real-time monitoring of cattle behavior is essential for enabling data-driven decision-making in precision livestock farming. However, existing monitoring solutions often rely on cloud-based processing or high-power hardware, which are impractical for deployment in remote or low-infrastructure agricultural environments. There is a critical need for low-cost, energy-efficient, and autonomous sensing systems capable of operating independently at the edge. This paper presents a compact, sensor-integrated system for real-time cattle behavior monitoring using an embedded vision sensor and a TinyML-based inference pipeline. The system is designed for low-power deployment in field conditions and integrates the OV2640 image sensor with the Sipeed Maixduino platform, which features the Kendryte K210 RISC-V processor and an on-chip neural network accelerator (KPU). The platform supports fully on-device classification of cattle postures using a quantized convolutional neural network trained on the publicly available cattle behavior dataset, covering standing and lying behavioral states. Sensor data is captured via the onboard camera and preprocessed in real time to meet model input specifications. The trained model is quantized and converted into a K210-compatible. kmodel using the NNCase toolchain, and deployed using MaixPy firmware. System performance was evaluated based on inference latency, classification accuracy, memory usage, and energy efficiency. Results demonstrate that the proposed TinyML-enabled system can accurately classify cattle behaviors in real time while operating within the constraints of a low-power, embedded platform, making it a viable solution for smart livestock monitoring in remote or under-resourced environments.

This study presents an incremental learning framework to enhance the generalization and robustness of transformer-based deep learning models for segmenting skin cancer and related tissue structures. While deep learning models often perform well on data distributions similar to their training sets, their accuracy typically degrades when exposed to novel scenarios—limiting their clinical utility in skin cancer diagnosis. To address this, we propose a biologically inspired incremental learning strategy tailored for skin cancer classification and segmentation, allowing the model to incorporate new data progressively while reducing catastrophic forgetting. Our approach integrates multiple loss functions to preserve existing knowledge while adapting to additional magnification levels. Experimental results on the in-distribution test set demonstrate consistent performance improvements: achieving 89.05% accuracy with 10× magnification, 92.68% with 10× and 5× combined, and 95.53% when incorporating 10×, 5×, and 2× magnifications. These findings highlight the potential of our method to improve the adaptability and reliability of deep learning systems for empirical generalization in skin cancer classification tasks. Please note Abbreviations should be introduced at the first mention in the main text – no abbreviations lists. Suggested structure of main text (not enforced) is provided below.

Software-based artificial life methods use mathematical and computational models to mimic complexity in living systems. Although such methods seem promising pertaining to exploring emergent behaviour, obtaining the governing rulesets of such methods remains challenging. In this paper, we present a concept of combined use of methods targeting different levels/scales in an emergent behaviour to obtain software-based artificial life rulesets from observed data. Additionally, we investigate the consequences of using this combination of methods by proposing an instance of combining Cellular Automata and Agent-based modelling with Koopman-based linearization. Our experiments on systems of Elementary Cellular Automaton (Rule 30), Game of Life (GOL), and Vicsek’s flocking show that the combined method can learn the overall non-linear and emergent behaviour, and the underlying governing rulesets. Our research also indicates that by identifying several emergent scales or levels in a system, the combined method has the potential to shed light on the learnt system dynamics.

'Retraction Note to: Toward Intelligent Telepresence Robotics for Enhancing Elderly Healthcare in Smart Care Home' published in 'Human-Computer Interaction'

Software-based artificial life methods use mathematical and computational models to mimic complexity in living systems. Although such methods seem promising pertaining to exploring emergent behaviour, obtaining the governing rulesets of such methods remains challenging. In this paper, we present a concept of combined use of methods targeting different levels/scales in an emergent behaviour to obtain software-based artificial life rulesets from observed data. Additionally, we investigate the consequences of using this combination of methods by proposing an instance of combining Cellular Automata (CA) and Agent-based modelling (ABM) with Koopman-based linearization. Our experiments on systems of Elementary Cellular Automaton (Rule 30), Game of Life (GOL), and Vicsek’s flocking show that the combined method can learn the overall non-linear and emergent behaviour, and the underlying governing rulesets. Our research also indicates that by identifying several emergent scales or levels in a system, the combined method has the potential to shed light on the learnt system dynamics.

Internet of Things (IoT) sensors designed for environmental and agricultural purposes can offer significant contributions to creating a sustainable and green environment. However, powering these sensors remains a challenge, and exploiting the temperature difference between air and soil appears to be a promising solution. For energy-harvesting technologies, accurate soil temperature profile data are needed. This study uses meteorological and soil temperature profile data collected in the Czech Republic to train machine learning models based on Polynomial Regression (PR), Support Vector Regression (SVR), and Long Short-Term Memory (LSTM) to predict the soil temperature profile. The results of the study indicate an error of 0.79 °C, which is approximately 10.9% lower than the temperature error reported in state-of-the-art studies. Beyond achieving a lower temperature prediction error, the proposed solution simplifies the input parameters of the model to only ambient temperature and solar irradiance. This improvement significantly reduces the computational costs associated with the regression model, offering a more efficient approach to predicting soil temperature for the purpose of optimizing energy harvesting in IoT sensors.