Kingsley A. Ogudo
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Kingsley A. Ogudo has a total of 57 co-authors. Here are the top co-authors by number of shared publications:
- Akintunde Ayodeji Alonge
Electrical and Electronic Engineering Technology, University of Johannesburg, Johannesburg, South Africa
10 shared publications - Ayodele Abiola Periola
Electrical, Electronic and Computer Engineering Department, Cape Peninsula University of Technology, Cape Town 7530, South Africa
9 shared publications - Lukas Willem Snyman
Department of Electrical Engineering, College of Science, Engineering and Technology, University of South Africa, Florida Campus, Johannesburg 1710, South Africa
9 shared publications Carlos Viana
Laboratoire Electronique, Systèmes de Communications et Microsystèmes, University Paris-Est, École Supérieure d’Ingénieurs en Électronique et Électrotechnique Paris Conservatoire National des Arts et Métiers, Université Paris-Est Marne-la-Vallée, Noisy-Le-Grand, France
4 shared publications- Joseph Tafataona Mtetwa
Department of Electrical and Electronics Engineering, University of Johannesburg, Johannesburg 2006, South Africa
4 shared publications - Sameerchand Pudaruth
ICT Department, University of Mauritius, Reduit 80837, Mauritius
4 shared publications
Latest Publications
- Open Access
- 11 November 2026
- Published by MDPI AG in World Electric Vehicle Journal
- Vol. 17 (11) , 591
Artificial Intelligence-Assisted Multi-Objective Design and Dynamic Validation of High-Efficiency Solid-State Transformers for Electric Vehicle Integration
Integrating electric vehicles (EVs) into power grids requires efficient, compact high-frequency transformers (HFTs) for solid-state-transformers (SST) and charging stages. This paper combines a physics-informed multi-objective genetic algorithm (NSGA-II, an evolutionary AI method), two-dimensional finite-element (FE) evaluation and dual-active-bridge (DAB) simulation to size HFTs at 85 kHz for ferrite 3C94 (Fe), Metglas 2605SA1 (Me) and Vitroperm 500F (Vi). In a first, FEMM-based stage (10 kW), analytical power density differed from FEMM by 73.7% (Fe), 22.7% (Vi) and 11.4% (Me); Vi reached 11.84 W/cm3 at a simulated peak transformer-only efficiency of 99.46%, and Litz optimisation identified AWG 44–46 and AWG 38–42 operating points. A second, independent stage (22 kW, five seeds per material) shows that the FE model raises the total loss by 3–11%, lowers the leakage inductance by about 48% and reveals hotspots about four times the lumped rise, excluding 31–54% of the designs at a 100 K limit. Fe and Vi then reach 6–37 and 7–27 W/cm3 at 99.8–99.9% efficiency and Me 5–9 W/cm3, and a designed DAB current loop is well damped (55° phase margin). The ranking is sensitive to loss fits and cooling prediction.
- 11 November 2026
- Published by MDPI AG in World Electric Vehicle Journal
- Vol. 17 (11) , 591
Open AccessArtificial Intelligence-Assisted Multi-Objective Design and Dynamic Validation of High-Efficiency Solid-State Transformers for Electric Vehicle Integration
Integrating electric vehicles (EVs) into power grids requires efficient, compact high-frequency transformers (HFTs) for solid-state-transformers (SST) and charging stages. This paper combines a physics-informed multi-objective genetic algorithm (NSGA-II, an evolutionary AI method), two-dimensional finite-element (FE) evaluation and dual-active-bridge (DAB) simulation to size HFTs at 85 kHz for ferrite 3C94 (Fe), Metglas 2605SA1 (Me) and Vitroperm 500F (Vi). In a first, FEMM-based stage (10 kW), analytical power density differed from FEMM by 73.7% (Fe), 22.7% (Vi) and 11.4% (Me); Vi reached 11.84 W/cm3 at a simulated peak transformer-only efficiency of 99.46%, and Litz optimisation identified AWG 44–46 and AWG 38–42 operating points. A second, independent stage (22 kW, five seeds per material) shows that the FE model raises the total loss by 3–11%, lowers the leakage inductance by about 48% and reveals hotspots about four times the lumped rise, excluding 31–54% of the designs at a 100 K limit. Fe and Vi then reach 6–37 and 7–27 W/cm3 at 99.8–99.9% efficiency and Me 5–9 W/cm3, and a designed DAB current loop is well damped (55° phase margin). The ranking is sensitive to loss fits and cooling prediction.
- Open Access
- 28 October 2026
- Published by MDPI AG in Automation
- Vol. 7 (6) , 174
Disturbance Rejection Technique on an Industrial Vacuum Furnace: The Model-Free iPID-DOB Approach
Slow non-minimum-phase (NMP) processes present significant control challenges due to inverse response behaviour, sluggish dynamics, and sensitivity to disturbances and model uncertainty. This study investigates the performance of two model-independent control strategies, namely a disturbance observer-based intelligent proportional–integral controller (iPID-DOB) and a Particle Swarm Optimization-tuned Active Disturbance Rejection Controller (ADRC-PSO), for the regulation of a slow NMP process identified from experimental open-loop data of an industrial vacuum furnace. The iPID-DOB integrates a conventional PI structure with an online disturbance observer to estimate and compensate disturbance-related effects, whereas the ADRC employs an extended state observer (ESO) optimized using Particle Swarm Optimization to estimate system states and lumped uncertainties in real time. Both controllers were evaluated under identical conditions involving multiple reference changes, actuator constraints, and repeated disturbance scenarios. Performance was assessed using steady-state error, root-mean-square error (RMSE), rise time, settling time, overshoot, integral error indices (ISE, IAE, and ITAE), control effort, disturbance-rejection capability, and recovery time. The results demonstrate that both controllers achieved stable closed-loop operation and effective disturbance attenuation across the investigated operating range. However, the iPID-DOB consistently outperformed the ADRC-PSO in all quantitative performance metrics, exhibiting faster convergence, lower tracking error, improved disturbance rejection, and reduced recovery times. Conversely, the ADRC-PSO generated a significantly smoother and more stable control action, primarily due to the optimized extended state observer, which provided accurate estimation of process states and lumped disturbances while suppressing excessive control activity. The findings indicate that the iPID-DOB offers superior tracking and robustness performance for the considered slow NMP process, whereas the ADRC-PSO provides enhanced control smoothness and operational stability. These results highlight the trade-off between aggressive disturbance-compensation strategies and observer-driven robust control, providing practical guidance for selecting suitable model-independent control architectures for challenging industrial processes.
- 28 October 2026
- Published by MDPI AG in Automation
- Vol. 7 (6) , 174
Open AccessDisturbance Rejection Technique on an Industrial Vacuum Furnace: The Model-Free iPID-DOB Approach
Slow non-minimum-phase (NMP) processes present significant control challenges due to inverse response behaviour, sluggish dynamics, and sensitivity to disturbances and model uncertainty. This study investigates the performance of two model-independent control strategies, namely a disturbance observer-based intelligent proportional–integral controller (iPID-DOB) and a Particle Swarm Optimization-tuned Active Disturbance Rejection Controller (ADRC-PSO), for the regulation of a slow NMP process identified from experimental open-loop data of an industrial vacuum furnace. The iPID-DOB integrates a conventional PI structure with an online disturbance observer to estimate and compensate disturbance-related effects, whereas the ADRC employs an extended state observer (ESO) optimized using Particle Swarm Optimization to estimate system states and lumped uncertainties in real time. Both controllers were evaluated under identical conditions involving multiple reference changes, actuator constraints, and repeated disturbance scenarios. Performance was assessed using steady-state error, root-mean-square error (RMSE), rise time, settling time, overshoot, integral error indices (ISE, IAE, and ITAE), control effort, disturbance-rejection capability, and recovery time. The results demonstrate that both controllers achieved stable closed-loop operation and effective disturbance attenuation across the investigated operating range. However, the iPID-DOB consistently outperformed the ADRC-PSO in all quantitative performance metrics, exhibiting faster convergence, lower tracking error, improved disturbance rejection, and reduced recovery times. Conversely, the ADRC-PSO generated a significantly smoother and more stable control action, primarily due to the optimized extended state observer, which provided accurate estimation of process states and lumped disturbances while suppressing excessive control activity. The findings indicate that the iPID-DOB offers superior tracking and robustness performance for the considered slow NMP process, whereas the ADRC-PSO provides enhanced control smoothness and operational stability. These results highlight the trade-off between aggressive disturbance-compensation strategies and observer-driven robust control, providing practical guidance for selecting suitable model-independent control architectures for challenging industrial processes.
- Open Access
- 7 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (17) , 8892
SwinIrisNet: A Hybrid Deep Learning Framework for Robust Iris Segmentation
Accurate iris segmentation remains a fundamental challenge in iris biometric recognition and medical image analysis, particularly in challenging scenarios such as non-cooperative acquisition conditions involving variable illumination, partial occlusions, degraded image quality, and diverse unconstrained environments. Prevailing segmentation algorithms exhibit limited robustness when confronted with such challenges, and the disparity between near-infrared (NIR) and visible-light imaging modalities further compounds the complexity of achieving a robust segmentation outcome. To address these challenges, this paper introduces SwinIrisNet, a hybrid deep learning architecture that integrates Swin Transformer and convolutional neural network (CNN) branches within a U-Net framework for robust iris segmentation. The Swin Transformer branch leverages hierarchical window-based self-attention to capture global contextual dependencies, whereas the CNN branch extracts fine-grained local features essential for precise boundary delineation. A memory-efficient cross-attention fusion module combines these complementary feature representations, further enhanced by a Convolutional Block Attention Module (CBAM), Atrous Spatial Pyramid Pooling (ASPP), and attention-gated skip connections for multi-scale context aggregation. An extensive evaluation is conducted across four publicly available benchmark datasets, including UBIRIS.v2, IITD, CASIA-Thousand, and MMU.v1, encompassing both visible-light and NIR imaging environments. The proposed architecture yields F1 values of 0.9612–0.9672, Dice coefficients of 0.9489–0.9519, mIoU values of 0.9266–0.9450, precision values of 0.9565–0.9633, recall values of 0.9600–0.9672, and classification accuracies of 99.51–99.53%, with NICE1 error rates of 0.57–0.60% and NICE2 values of 1.82–2.24%, confirming pixel-level segmentation quality. Cross-database generalization experiments further demonstrate that SwinIrisNet learns transferable iris representations and generalizes effectively across heterogeneous imaging sources, with the strongest transfer occurring in the NIR-to-visible direction. A comparative analysis against existing algorithms demonstrates that the proposed architecture attains substantial performance improvements over several existing segmentation networks when evaluated on identical benchmark databases, surpassing them across the majority of qualitative and quantitative metrics while maintaining a marginally lower memory footprint.
- 7 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (17) , 8892
Open AccessSwinIrisNet: A Hybrid Deep Learning Framework for Robust Iris Segmentation
Accurate iris segmentation remains a fundamental challenge in iris biometric recognition and medical image analysis, particularly in challenging scenarios such as non-cooperative acquisition conditions involving variable illumination, partial occlusions, degraded image quality, and diverse unconstrained environments. Prevailing segmentation algorithms exhibit limited robustness when confronted with such challenges, and the disparity between near-infrared (NIR) and visible-light imaging modalities further compounds the complexity of achieving a robust segmentation outcome. To address these challenges, this paper introduces SwinIrisNet, a hybrid deep learning architecture that integrates Swin Transformer and convolutional neural network (CNN) branches within a U-Net framework for robust iris segmentation. The Swin Transformer branch leverages hierarchical window-based self-attention to capture global contextual dependencies, whereas the CNN branch extracts fine-grained local features essential for precise boundary delineation. A memory-efficient cross-attention fusion module combines these complementary feature representations, further enhanced by a Convolutional Block Attention Module (CBAM), Atrous Spatial Pyramid Pooling (ASPP), and attention-gated skip connections for multi-scale context aggregation. An extensive evaluation is conducted across four publicly available benchmark datasets, including UBIRIS.v2, IITD, CASIA-Thousand, and MMU.v1, encompassing both visible-light and NIR imaging environments. The proposed architecture yields F1 values of 0.9612–0.9672, Dice coefficients of 0.9489–0.9519, mIoU values of 0.9266–0.9450, precision values of 0.9565–0.9633, recall values of 0.9600–0.9672, and classification accuracies of 99.51–99.53%, with NICE1 error rates of 0.57–0.60% and NICE2 values of 1.82–2.24%, confirming pixel-level segmentation quality. Cross-database generalization experiments further demonstrate that SwinIrisNet learns transferable iris representations and generalizes effectively across heterogeneous imaging sources, with the strongest transfer occurring in the NIR-to-visible direction. A comparative analysis against existing algorithms demonstrates that the proposed architecture attains substantial performance improvements over several existing segmentation networks when evaluated on identical benchmark databases, surpassing them across the majority of qualitative and quantitative metrics while maintaining a marginally lower memory footprint.
- Open Access
- 20 July 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (14) , 2637
Curvature-Corrected Rotary Position Embeddings: An Entropy-Invariant Temperature for Mitigating Numerical-Rank Collapse in Long-Context Attention
As large language models process ever longer contexts, the positional encoding—most commonly rotary position embeddings (RoPEs)—must remain numerically trustworthy. We give an information-geometric analysis of why the attention matrix becomes ill-conditioned as the sequence length L grows. We first show that the tempting frequency-aliasing explanation fails, because RoPE’s multi-frequency code keeps positions well separated. The mechanism is statistical: at fixed softmax temperature the attention entropy grows like , and the rows spread their mass over a collision support (inverse participation ratio) that grows polynomially as ( measured), confining the matrix’s trailing singular values and producing a numerical-rank collapse. We prove, and confirm in 50-digit arithmetic, that the spectral condition number is intrinsically ill-posed here, so the numerical rank is the correct diagnostic. From the exact entropy–temperature identity we derive the entropy-invariant schedule , Curvature-Corrected RoPE (CC-RoPE), and prove that in the diffuse near-circulant regime it eliminates the collapse at a characterised conditioning–mixing cost, whereas no relative-position-preserving warp can. The two signatures are further confirmed in a trained RoPE model (Pythia-160M); the downstream perplexity effect is framed as a falsifiable prediction.- 20 July 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (14) , 2637
Open AccessCurvature-Corrected Rotary Position Embeddings: An Entropy-Invariant Temperature for Mitigating Numerical-Rank Collapse in Long-Context Attention
As large language models process ever longer contexts, the positional encoding—most commonly rotary position embeddings (RoPEs)—must remain numerically trustworthy. We give an information-geometric analysis of why the attention matrix becomes ill-conditioned as the sequence length L grows. We first show that the tempting frequency-aliasing explanation fails, because RoPE’s multi-frequency code keeps positions well separated. The mechanism is statistical: at fixed softmax temperature the attention entropy grows like , and the rows spread their mass over a collision support (inverse participation ratio) that grows polynomially as ( measured), confining the matrix’s trailing singular values and producing a numerical-rank collapse. We prove, and confirm in 50-digit arithmetic, that the spectral condition number is intrinsically ill-posed here, so the numerical rank is the correct diagnostic. From the exact entropy–temperature identity we derive the entropy-invariant schedule , Curvature-Corrected RoPE (CC-RoPE), and prove that in the diffuse near-circulant regime it eliminates the collapse at a characterised conditioning–mixing cost, whereas no relative-position-preserving warp can. The two signatures are further confirmed in a trained RoPE model (Pythia-160M); the downstream perplexity effect is framed as a falsifiable prediction. - Open Access
- 22 May 2026
- Published by MDPI AG in Automation
- Vol. 7 (3) , 84
Hybrid Mamdani–ANFIS Data-Driven Control on an Industrial Heating Furnace
The research presented provides an overview of the latest progress in data-driven control methods used for industrial heating furnaces. Although the data-driven methodologies reviewed provide good performance metrics compared to conventional control strategies, they lack the integration of energy efficiency considerations into the controller design process. This research presents a comprehensive control design framework for a novel energy-efficient data-driven controller applied to an industrial heating furnace. It proposes a novel Hybrid Mamdani–ANFIS controller developed using real-time data from an industrial heating furnace. A novel ANFIS-based energy model is also presented in this work to evaluate the energy efficiency of the presented controller models. The results demonstrated that the proposed novel Hybrid Mamdani–ANFIS controller outperforms both the Fuzzy PID and conventional Fuzzy controller in terms of energy efficiency, achieving approximately 30% energy savings and exhibiting a faster disturbance response time. This study makes a considerable contribution to the field of control theory by synthesizing existing knowledge, addressing identified research gaps, and introducing a novel control design framework that enhances energy efficiency, robustness, and adaptability across a wide spectrum of control applications in industrial heating furnace systems.
- 22 May 2026
- Published by MDPI AG in Automation
- Vol. 7 (3) , 84
Open AccessHybrid Mamdani–ANFIS Data-Driven Control on an Industrial Heating Furnace
The research presented provides an overview of the latest progress in data-driven control methods used for industrial heating furnaces. Although the data-driven methodologies reviewed provide good performance metrics compared to conventional control strategies, they lack the integration of energy efficiency considerations into the controller design process. This research presents a comprehensive control design framework for a novel energy-efficient data-driven controller applied to an industrial heating furnace. It proposes a novel Hybrid Mamdani–ANFIS controller developed using real-time data from an industrial heating furnace. A novel ANFIS-based energy model is also presented in this work to evaluate the energy efficiency of the presented controller models. The results demonstrated that the proposed novel Hybrid Mamdani–ANFIS controller outperforms both the Fuzzy PID and conventional Fuzzy controller in terms of energy efficiency, achieving approximately 30% energy savings and exhibiting a faster disturbance response time. This study makes a considerable contribution to the field of control theory by synthesizing existing knowledge, addressing identified research gaps, and introducing a novel control design framework that enhances energy efficiency, robustness, and adaptability across a wide spectrum of control applications in industrial heating furnace systems.
- Open Access
A Validation-Driven Explainable Deep Ensemble Framework for Image-Based Saffron Adulteration Detection
Saffron (Crocus sativus L.), one of the world’s most valuable spices, is highly vulnerable to adulteration due to its premium market price and the limitations of conventional analytical methods for rapid, non-destructive authentication. Although recent deep learning-based approaches have reported promising accuracy, many rely on single models or naïve ensembles and lack rigorous validation and statistical reliability assessment. This study proposes a validation-driven and explainable deep ensemble framework for image-based saffron adulteration detection. Multiple pretrained convolutional neural networks (DenseNet169, ResNet50, and VGG16) are integrated using a validation-driven weighted ensemble strategy in which fusion weights are computed exclusively from validation performance within the training folds and fixed prior to evaluation on the held-out fold, thereby preventing information leakage between model selection and performance assessment. The proposed framework achieved 98.61% classification accuracy, 98.17% F1-score, and 98.61% AUC, outperforming the best individual base model by up to 1.4% in F1-score. Stratified five-fold cross-validation demonstrated stable performance, with a mean accuracy of 97.81% ± 0.53, confirming robustness across data splits. Statistical validation using McNemar’s test (p < 0.05) and 5 × 2 cross-validated significance testing verified that performance improvements over constituent models are statistically reliable. Grad-CAM-based explainability and background-invariance analysis further demonstrated that predictions are driven primarily by intrinsic filament-level characteristics, with only a marginal (~0.9%) performance reduction under ROI-cropped evaluation. The proposed framework provides a robust, interpretable, and statistically validated solution for saffron authentication and offers methodological insights for reliable image-based food adulteration detection under limited data conditions.
Open AccessA Validation-Driven Explainable Deep Ensemble Framework for Image-Based Saffron Adulteration Detection
Saffron (Crocus sativus L.), one of the world’s most valuable spices, is highly vulnerable to adulteration due to its premium market price and the limitations of conventional analytical methods for rapid, non-destructive authentication. Although recent deep learning-based approaches have reported promising accuracy, many rely on single models or naïve ensembles and lack rigorous validation and statistical reliability assessment. This study proposes a validation-driven and explainable deep ensemble framework for image-based saffron adulteration detection. Multiple pretrained convolutional neural networks (DenseNet169, ResNet50, and VGG16) are integrated using a validation-driven weighted ensemble strategy in which fusion weights are computed exclusively from validation performance within the training folds and fixed prior to evaluation on the held-out fold, thereby preventing information leakage between model selection and performance assessment. The proposed framework achieved 98.61% classification accuracy, 98.17% F1-score, and 98.61% AUC, outperforming the best individual base model by up to 1.4% in F1-score. Stratified five-fold cross-validation demonstrated stable performance, with a mean accuracy of 97.81% ± 0.53, confirming robustness across data splits. Statistical validation using McNemar’s test (p < 0.05) and 5 × 2 cross-validated significance testing verified that performance improvements over constituent models are statistically reliable. Grad-CAM-based explainability and background-invariance analysis further demonstrated that predictions are driven primarily by intrinsic filament-level characteristics, with only a marginal (~0.9%) performance reduction under ROI-cropped evaluation. The proposed framework provides a robust, interpretable, and statistically validated solution for saffron authentication and offers methodological insights for reliable image-based food adulteration detection under limited data conditions.
- Open Access
- 6 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 53019-53037
AD-ViTNet: Modeling Global Structural Dependencies in Brain MRI for Alzheimer’s Disease Detection Using Vision Transformer
Although convolutional neural networks (CNNs) have shown promising results for automated detection Alzheimer’s disease (AD) from structural MRI, their reliance on local receptive fields limits their ability to capture the diffuse and spatially distributed neurodegenerative changes characteristic of the disease. To address this limitation, this study proposes the AD-ViTNet, a Vision Transformer-based framework to model global structural dependencies in brain MRI through self-attention mechanisms. The proposed AD-ViTNet framework trained and evaluated on structural MRI data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) processes MRI scans using patch-wise representations and multi-head self-attention to learn global contextual features across the entire brain volume. Comprehensive comparative experiments are conducted against widely used CNN architectures, including ResNet50, EfficientNetB0, InceptionNet, and VGG16, for binary classification between cognitively normal control subjects and individuals diagnosed with AD. Experimental results show that the AD-ViTNet consistently outperforms all benchmark models across multiple evaluation metrics. In the primary evaluation setting, AD-ViTNet achieves an accuracy of 96.8%, F1-score of 96.8%, and AUC of 0.983, indicating strong discriminative capability and robust feature representation. Statistical significance is verified using McNemar’s test, with all comparisons yielding p-values < 0.05, confirming that the observed performance improvements are not attributable to chance. These findings demonstrate the effectiveness of global attention mechanisms in capturing heterogeneous atrophy patterns associated with AD and highlight the potential of transformer-based architectures for reliable, computer-aided neuroimaging diagnosis.
- 6 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 53019-53037
Open AccessAD-ViTNet: Modeling Global Structural Dependencies in Brain MRI for Alzheimer’s Disease Detection Using Vision Transformer
Although convolutional neural networks (CNNs) have shown promising results for automated detection Alzheimer’s disease (AD) from structural MRI, their reliance on local receptive fields limits their ability to capture the diffuse and spatially distributed neurodegenerative changes characteristic of the disease. To address this limitation, this study proposes the AD-ViTNet, a Vision Transformer-based framework to model global structural dependencies in brain MRI through self-attention mechanisms. The proposed AD-ViTNet framework trained and evaluated on structural MRI data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) processes MRI scans using patch-wise representations and multi-head self-attention to learn global contextual features across the entire brain volume. Comprehensive comparative experiments are conducted against widely used CNN architectures, including ResNet50, EfficientNetB0, InceptionNet, and VGG16, for binary classification between cognitively normal control subjects and individuals diagnosed with AD. Experimental results show that the AD-ViTNet consistently outperforms all benchmark models across multiple evaluation metrics. In the primary evaluation setting, AD-ViTNet achieves an accuracy of 96.8%, F1-score of 96.8%, and AUC of 0.983, indicating strong discriminative capability and robust feature representation. Statistical significance is verified using McNemar’s test, with all comparisons yielding p-values < 0.05, confirming that the observed performance improvements are not attributable to chance. These findings demonstrate the effectiveness of global attention mechanisms in capturing heterogeneous atrophy patterns associated with AD and highlight the potential of transformer-based architectures for reliable, computer-aided neuroimaging diagnosis.
- 3 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Development of edge-emitting Si/SiGe based optical sources operating in the visible and near visible range wavelength for sensing and communication applications
Développement d'optiques à base de Si / SiGe sources opérant dans le visible et le proche visible longueur d'onde pour les applications de détection et de communication: Nous proposons des liaisons optiques en Silicium à faible coût utilisant des longueurs d'onde de propagation de 650 à 850 nm. La création de circuits intégrés optoélectroniques à grande échelle et de bus de données optiques au sein même des circuits intégrés, utilisant des composés Silicium CMOS, ont été envisagées présentant une voie prometteuse [1] - [3]. Dans les dernières tentatives de réalisation de systèmes optoélectroniques en CMOS, les technologies était principalement orientée sur l'utilisation des longueurs d'onde à 1550 nm [4] - [6], principalement en raison de la facilité de conception et de fabrication des guides d'ondes dans ce régime de longueur d'onde. Cependant, aucune source optique rapide efficace et aucun photo-détecteur Si ne sont disponibles à cette longueur d'onde de 1550 nm. Aujourd'hui, les solutions pour surmonter le problème sont principalement axées sur l'intégration de sources optiques basées sur des éléments du groupe IIIV reportés sur Silicium par liaison moléculaire [7a] - [7b].Si la source optique, le détecteur, les guides d'ondes et les capteurs pouvaient être réalisés sur la même puce CMOS Silicium, par exemple à une longueur d'onde de 750 nm, divers systèmes micro-photoniques sur puces, légers et miniaturisés, pourraient être conçus et réalisés. Alors que les sources optiques au Silicium ne sont peut-être pas encore au niveau de performance requis pour les communications à très haut débit, les systèmes optoélectroniques "tout-Silicium" à faible coût restent encore un excellent point de départ. Ces sources pourraient également conduire à un nouveau champ qui pourrait s'appeler «microsystèmes photoniques Si» ouvrant la voie à de nouvelles applications et produits notamment pour l'optique médicale, biomédicale, les interconnexions optiques et la biophotonique. Ces systèmes ne nécessitent de bande passante à très haute fréquence pour émettre, et les puissances d'émission de nos diodes électroluminescentes (LED) à avalanche peuvent être suffisantes pour assurer le fonctionnement de tels systèmes. Ce travail de thèse de doctorat traite donc des liaisons optiques SiGe / Si à faible coût en utilisant des dispositifs Photonique-Microondes tels que une source à Diode Electro-Luminescente (DEL) à avalanche en Silicium (SiAvLED) et Silicium-Germanium intégrée en technologie bipolaire, des guides d'ondes optiques en Nitrure de Silicium et en Oxyde de Silicium, des phototransistors bipolaires à hétérojonction (HPT) SiGe. Ce travail se concentre sur l'intégration combinée de sources optiques à l'échelle micrométrique, de guides d'ondes optiques et de détecteurs sur une même puce pour former une liaison de communication complète pour diverses applications iv impliquant des liaisons de courte longueur d'onde (750 nm à 950 nm). Les progrès fournis par ce travail par rapport aux travaux antérieurs pourraient être synthétisés comme suit:• La source optique, le guide d'ondes et le détecteur ont tous été intégrés et alignés sur la même puce, dans une technologie industrielle RF bipolaire SiGe 0,35μm, pour former une liaison optique ou optique micro-onde complète sur puce à la longueur d'onde de 750 nm.• Une série de liaisons de communication optique de deuxième génération de 50μm de longueur, utilisant des longueurs d'onde de propagation de 650 à 850 nm, a été conçue et réalisée en technologie SiGe bipolaire également. Des sources optiques, des guides d'ondes et des détecteurs de dimensions micrométriques ont tous été intégrés sur la même puce pour former une communication complète sur les liaisons micro-optiques. Des LED Si à base d'avalanche (Si Av LED), des contacts Schottky, des stratégies de densification TEOS, des guides d'ondes à base de Nitrure de Silicium et des technologies de détection bipolaire SiGe de pointe ont été utilisées comme stratégies de conception clés.• Le logiciel de simulation R-soft (Beam Prop) a été utilisé comme outil de simulation
- 3 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Development of edge-emitting Si/SiGe based optical sources operating in the visible and near visible range wavelength for sensing and communication applications
Développement d'optiques à base de Si / SiGe sources opérant dans le visible et le proche visible longueur d'onde pour les applications de détection et de communication: Nous proposons des liaisons optiques en Silicium à faible coût utilisant des longueurs d'onde de propagation de 650 à 850 nm. La création de circuits intégrés optoélectroniques à grande échelle et de bus de données optiques au sein même des circuits intégrés, utilisant des composés Silicium CMOS, ont été envisagées présentant une voie prometteuse [1] - [3]. Dans les dernières tentatives de réalisation de systèmes optoélectroniques en CMOS, les technologies était principalement orientée sur l'utilisation des longueurs d'onde à 1550 nm [4] - [6], principalement en raison de la facilité de conception et de fabrication des guides d'ondes dans ce régime de longueur d'onde. Cependant, aucune source optique rapide efficace et aucun photo-détecteur Si ne sont disponibles à cette longueur d'onde de 1550 nm. Aujourd'hui, les solutions pour surmonter le problème sont principalement axées sur l'intégration de sources optiques basées sur des éléments du groupe IIIV reportés sur Silicium par liaison moléculaire [7a] - [7b].Si la source optique, le détecteur, les guides d'ondes et les capteurs pouvaient être réalisés sur la même puce CMOS Silicium, par exemple à une longueur d'onde de 750 nm, divers systèmes micro-photoniques sur puces, légers et miniaturisés, pourraient être conçus et réalisés. Alors que les sources optiques au Silicium ne sont peut-être pas encore au niveau de performance requis pour les communications à très haut débit, les systèmes optoélectroniques "tout-Silicium" à faible coût restent encore un excellent point de départ. Ces sources pourraient également conduire à un nouveau champ qui pourrait s'appeler «microsystèmes photoniques Si» ouvrant la voie à de nouvelles applications et produits notamment pour l'optique médicale, biomédicale, les interconnexions optiques et la biophotonique. Ces systèmes ne nécessitent de bande passante à très haute fréquence pour émettre, et les puissances d'émission de nos diodes électroluminescentes (LED) à avalanche peuvent être suffisantes pour assurer le fonctionnement de tels systèmes. Ce travail de thèse de doctorat traite donc des liaisons optiques SiGe / Si à faible coût en utilisant des dispositifs Photonique-Microondes tels que une source à Diode Electro-Luminescente (DEL) à avalanche en Silicium (SiAvLED) et Silicium-Germanium intégrée en technologie bipolaire, des guides d'ondes optiques en Nitrure de Silicium et en Oxyde de Silicium, des phototransistors bipolaires à hétérojonction (HPT) SiGe. Ce travail se concentre sur l'intégration combinée de sources optiques à l'échelle micrométrique, de guides d'ondes optiques et de détecteurs sur une même puce pour former une liaison de communication complète pour diverses applications iv impliquant des liaisons de courte longueur d'onde (750 nm à 950 nm). Les progrès fournis par ce travail par rapport aux travaux antérieurs pourraient être synthétisés comme suit:• La source optique, le guide d'ondes et le détecteur ont tous été intégrés et alignés sur la même puce, dans une technologie industrielle RF bipolaire SiGe 0,35μm, pour former une liaison optique ou optique micro-onde complète sur puce à la longueur d'onde de 750 nm.• Une série de liaisons de communication optique de deuxième génération de 50μm de longueur, utilisant des longueurs d'onde de propagation de 650 à 850 nm, a été conçue et réalisée en technologie SiGe bipolaire également. Des sources optiques, des guides d'ondes et des détecteurs de dimensions micrométriques ont tous été intégrés sur la même puce pour former une communication complète sur les liaisons micro-optiques. Des LED Si à base d'avalanche (Si Av LED), des contacts Schottky, des stratégies de densification TEOS, des guides d'ondes à base de Nitrure de Silicium et des technologies de détection bipolaire SiGe de pointe ont été utilisées comme stratégies de conception clés.• Le logiciel de simulation R-soft (Beam Prop) a été utilisé comme outil de simulation
- Open Access
Foundation Protein Language Models for Influenza A Virus T-Cell Epitope Prediction: A Transformer-Based Viroinformatics Framework
Influenza A virus remains a major cause of respiratory disease worldwide and poses a persistent challenge to vaccine development due to its rapid genetic evolution and antigenic variability. T-cell-based immunity has therefore gained increasing importance, as it can provide broader and more durable protection by targeting conserved viral regions. Accurate identification of T-cell epitopes (TCEs) is a fundamental requirement for epitope-based vaccine design and immunological research. Although numerous computational methods have been proposed, many existing approaches rely on handcrafted physicochemical features, which offer limited ability to capture contextual sequence dependencies. In this study, a transformer-based viroinformatics framework is proposed for the binary prediction of TCEs from Influenza A virus peptide sequences. The framework employs a pretrained Evolutionary Scale Modeling-2 (ESM-2) protein language model (PLM) to generate rich, contextualized embeddings directly from raw amino acid sequences, eliminating the need for manual feature engineering. These embeddings are processed using a lightweight attention-based transformer classifier to learn epitope-specific sequence patterns. The model achieves strong and stable predictive performance, attaining an accuracy of approximately 97% and an AUC close to 0.99 under stratified cross-validation. Ablation analysis further confirms that protein language model representations and self-attention contribute substantially to performance gains over classical machine learning baselines. To enhance practical reliability, Monte Carlo dropout is incorporated during inference to provide uncertainty-aware predictions, enabling differentiation between high-confidence and ambiguous peptide candidates. In addition, attention-based interpretability is used to identify residue-level contributions to model decisions, offering biologically meaningful insights into epitope recognition. Overall, this study demonstrates that PLMs combined with Transformer architectures provide an effective, interpretable, and a promising computational framework for Influenza A TCE discovery and vaccine research.
Open AccessFoundation Protein Language Models for Influenza A Virus T-Cell Epitope Prediction: A Transformer-Based Viroinformatics Framework
Influenza A virus remains a major cause of respiratory disease worldwide and poses a persistent challenge to vaccine development due to its rapid genetic evolution and antigenic variability. T-cell-based immunity has therefore gained increasing importance, as it can provide broader and more durable protection by targeting conserved viral regions. Accurate identification of T-cell epitopes (TCEs) is a fundamental requirement for epitope-based vaccine design and immunological research. Although numerous computational methods have been proposed, many existing approaches rely on handcrafted physicochemical features, which offer limited ability to capture contextual sequence dependencies. In this study, a transformer-based viroinformatics framework is proposed for the binary prediction of TCEs from Influenza A virus peptide sequences. The framework employs a pretrained Evolutionary Scale Modeling-2 (ESM-2) protein language model (PLM) to generate rich, contextualized embeddings directly from raw amino acid sequences, eliminating the need for manual feature engineering. These embeddings are processed using a lightweight attention-based transformer classifier to learn epitope-specific sequence patterns. The model achieves strong and stable predictive performance, attaining an accuracy of approximately 97% and an AUC close to 0.99 under stratified cross-validation. Ablation analysis further confirms that protein language model representations and self-attention contribute substantially to performance gains over classical machine learning baselines. To enhance practical reliability, Monte Carlo dropout is incorporated during inference to provide uncertainty-aware predictions, enabling differentiation between high-confidence and ambiguous peptide candidates. In addition, attention-based interpretability is used to identify residue-level contributions to model decisions, offering biologically meaningful insights into epitope recognition. Overall, this study demonstrates that PLMs combined with Transformer architectures provide an effective, interpretable, and a promising computational framework for Influenza A TCE discovery and vaccine research.
- Open Access
Constrained and Unconstrained Control Design of Electromagnetic Levitation System with Integral Robust–Optimal Sliding Mode Control for Mismatched Uncertainties
In real life, almost all systems are nonlinear in nature. The electromagnetic levitation system (EMLS) is one such system that has a wide range of applications due to its frictionless, fast, and affordable technique. Optimal control and sliding mode control (SMC) techniques are often used controllers for EMLS. However, these techniques can achieve the required levitation but lag in having perfect set-point tracking and robustness against uncertainties. To get over these drawbacks, this article proposes the design of unconstrained mismatched uncertainties, constrained mismatched uncertainties, and integral sliding mode control with mismatched uncertainties for the current-controlled-type electromagnetic levitation system (CC-EMLS). The modeled equations of CC-EMLS are transfomed in terms of the mismatched uncertainties, and the required control action is obtained with and without constraints on the control input. The quadratic performance function is suggested for the unconstrained control scheme and is solved using the Hamilton–Jacobi–Bellman (HJB) equation. The non-quadratic cost function is designed for the constrained control method, and the HJB equation is utilized to obtain the solution. Both control schemes provide robustness to the system, but deviations in the set point are observed in tracking the position of the ball when the changes in the payload occur in the system. Therefore, integral sliding mode control with robust–optimal (IOSMC) gain is proposed for the CC-EMLS to overcome the steady-state error in the other two schemes. The stability is proven using the direct method of Lyapunov stability. The essential studies based on the simulation are carried out to showcase the performance of the proposed control schemes. The integral performance indicators are compared for all three proposed control schemes to highlight the efficacy, robustness, and efficiency of the designed controllers.
Open AccessConstrained and Unconstrained Control Design of Electromagnetic Levitation System with Integral Robust–Optimal Sliding Mode Control for Mismatched Uncertainties
In real life, almost all systems are nonlinear in nature. The electromagnetic levitation system (EMLS) is one such system that has a wide range of applications due to its frictionless, fast, and affordable technique. Optimal control and sliding mode control (SMC) techniques are often used controllers for EMLS. However, these techniques can achieve the required levitation but lag in having perfect set-point tracking and robustness against uncertainties. To get over these drawbacks, this article proposes the design of unconstrained mismatched uncertainties, constrained mismatched uncertainties, and integral sliding mode control with mismatched uncertainties for the current-controlled-type electromagnetic levitation system (CC-EMLS). The modeled equations of CC-EMLS are transfomed in terms of the mismatched uncertainties, and the required control action is obtained with and without constraints on the control input. The quadratic performance function is suggested for the unconstrained control scheme and is solved using the Hamilton–Jacobi–Bellman (HJB) equation. The non-quadratic cost function is designed for the constrained control method, and the HJB equation is utilized to obtain the solution. Both control schemes provide robustness to the system, but deviations in the set point are observed in tracking the position of the ball when the changes in the payload occur in the system. Therefore, integral sliding mode control with robust–optimal (IOSMC) gain is proposed for the CC-EMLS to overcome the steady-state error in the other two schemes. The stability is proven using the direct method of Lyapunov stability. The essential studies based on the simulation are carried out to showcase the performance of the proposed control schemes. The integral performance indicators are compared for all three proposed control schemes to highlight the efficacy, robustness, and efficiency of the designed controllers.