PPGEE - Doutorado em Engenharia Elétrica e de Computação
Permanent URI for this collectionhttps://repositorio.ufrn.br/handle/123456789/12009
Browse
collection.page.browse.recent.head
Now showing 1 - 20 of 399
Doctoral Thesis Treinamento consciente de quantização vetorial evolutiva: um método adaptativo para compactar modelos de aprendizado de máquina(Universidade Federal do Rio Grande do Norte, 2026-02-06) Flores, Thommas Kevin Sales; Silva, Ivanovitch Medeiros Dantas da; Costa, Daniel Gouveia; https://orcid.org/0000-0003-3988-8476; http://lattes.cnpq.br/0200709054904002; https://orcid.org/0000-0002-0116-6489; http://lattes.cnpq.br/3608440944832201; https://orcid.org/0000-0003-2808-8529; http://lattes.cnpq.br/0630479458408181; Vieira, Tiago Figueiredo; http://lattes.cnpq.br/8601011832053651; Martins, Allan de Medeiros; https://orcid.org/0000-0002-9486-4509; http://lattes.cnpq.br/4402694969508077; Villanueva, Juan Moisés Mauricio; https://orcid.org/0000-0002-8760-9390; http://lattes.cnpq.br/1715030763709959; Oliveira, Luiz Affonso Henderson Guedes de; https://orcid.org/0000-0003-2690-1563; http://lattes.cnpq.br/7987212907837941The consolidation of Large Language Models poses significant challenges to their deployment on edge devices, as the reliance on cloud-based infrastructures compromises data privacy, increases latency, and limits critical applications in scenarios with unstable connectivity. Traditional compression techniques, predominantly applied after training, prove inadequate in this context, as they may degrade model performance or rely on opaque execution pipelines, hindering the analysis and verification of system behavior. In response to these challenges, this thesis proposes Evolving Vector Quantization-Aware Training (EVQAT), whose central principle consists of integrating compression directly into the learning process. In this approach, quantization is no longer treated as a post-training step but rather as an adaptive mechanism during training, allowing the transformations imposed on model parameters to be explicitly controlled and analyzed. mThe methodology is grounded in evolutionary vector quantization theory by treating the weights of the language model as continuous data streams, enabling incremental clustering algorithms such as AutoCloud, Mean-Shift, and Affinity Propagation to dynamically adjust the codebooks in synchrony with the optimization of high-dimensional parameters, thereby enabling model compression with explicit control over quantization error. The integration of codebook evolution into the training process ensures that the partitioning of the vector space follows the learning dynamics, preserving relevant statistical properties and reducing the impact of quantization noise. This characteristic makes EVQAT suitable for quantizationsensitive architectures such as Transformers, extending its applicability from multilayer neural networks to generative language models under strict memory and computational constraints. The proposed approach is evaluated across multiple experimental scenarios, including training without quantization, quantization-aware training using evolutionary vector quantization and 8-bit integer (INT8) representations, as well as post-training quantization under the same variants, with deployments on microcontroller-based platforms. In this context, auditability is achieved through transparent access to model parameters, including the vocabulary, the weights of each hidden layer, and the attention mechanisms. As a result, the methodology enables the automatic generation of standard, self-contained C++ code, with real-time inference capabilities and direct execution on diverse hardware architectures, without reliance on third-party libraries or operating systems, ensuring explainability at the levels of training, model, and execution, as well as interoperability across platforms.Doctoral Thesis Control of an automated insulin delivery system using artificial intelligence(Universidade Federal do Rio Grande do Norte, 2025-10-02) Farias, João Lucas Correia Barbosa de; Bessa, Wallace Moreira; https://orcid.org/0000-0002-0935-7730; http://lattes.cnpq.br/3256782908311485; https://orcid.org/0000-0002-3324-2239; http://lattes.cnpq.br/0575163993612915; Bertachi, Arthur Hirata; http://lattes.cnpq.br/2283634850801378; Cunha Júnior, Américo Barbosa da; https://orcid.org/0000-0002-8342-0363; http://lattes.cnpq.br/5659403706694491; Dórea, Carlos Eduardo Trabuco; https://orcid.org/0000-0002-3999-2874; http://lattes.cnpq.br/0143490577842914; Medeiros, Philippe Eduardo de; https://orcid.org/0000-0002-1803-241X; http://lattes.cnpq.br/5594617038786753Type 1 Diabetes Mellitus is a disease that affects millions of people worldwide. Recently, devices that automatically regulate blood glucose concentration in diabetic patients were developed. The Automated Insulin Delivery (AID) system can provide a life with higher quality, autonomy, and comfort for patients. This work aims to design an intelligent nonlinear controller with an estimator developed with Artificial Intelligence (AI) algorithms for an AID system. An extensive review on diabetes is conducted introducing its definition, physiological context, diagnosis, global impact and available state of the art therapies. Furthermore, a literature review is carried out presenting the types of AID systems, current challenges, virtual patient models, control strategies and AI algorithms that shape modern designs. The OHSU virtual patient model, used to simulate the virtual patient population in this work, is introduced along with its mathematical formulation, which incorporates physical activity to the insulin-glucose dynamics. The proposed controller combines Feedback Linearization with an intelligent neural network-based estimator to compensate for model uncertainties, external disturbances and time delays inherent of the biological system in question. It incorporates a Reinforcement Learning algorithm, Q-learning, to optimize neural network learning rate across a diverse virtual patient population while minimizing hypoglycemic events. Results show the controller was able to safely and effectively regulate glycemia with average Time in Range (TIR) of 81% and Coefficient of Variation (CV) of 35% on a diverse training population of 10 subjects, and average TIR of 76% and CV of 34% for a different unseen validation population of 5 subjects. In order to avoid poor control during the network’s initial learning period, offline network pre-training is proposed and tested on the validation population. Severe hyperglycemic events in the early days of simulation were completely avoided with the added benefit of a slight improvement in TIR (76% vs. 78%), indicating a clear path on how to introduce the controller to real-life subjects. Overall, the proposed controller was able to provide safe and effective glycemic control under exercising across a diverse virtual patient population without the need for user input on meal intakes or physical activities.Doctoral Thesis A Knowledge Graph-Based approach for modeling legislative texts: representation and document similarity analysis(Universidade Federal do Rio Grande do Norte, 2025-08-29) Oliveira, Gisliany Lillian Alves de; Silva, Ivanovitch Medeiros Dantas da; https://orcid.org/0000-0002-6037-9660; http://lattes.cnpq.br/8193679068029541; Viegas, Carlos Manuel Dias; Villanueva, Juan Moisés Mauricio; Oliveira, Luiz Affonso Henderson Guedes de; Barros, Thiago MedeirosThe most prominent task of the Legislative Branch — lawmaking — depends on a complex and demanding process in which new proposals must be examined, debated, and revised in light of existing legislation. These activities are often labor-intensive for humans due to the technical language, substantial length, and interdependence of legal texts. At the same time, these characteristics present a tangible opportunity for Artificial Intelligence (AI), particularly through the integration of Natural Language Processing (NLP) and structured data representations. Aiming to model legislative documents in a way that preserves their rich structural semantics, this work proposes an approach for transforming legislative texts into domain-specialized Knowledge Graphs (KGs) that capture their inherent hierarchical organization. The proposed method, based on LexML standards — a Brazilian XML schema for legal documents — extracts explicit structural relationships (e.g., articles, paragraphs, items) and augments them with contextual entities and relationships extracted by a Large Language Model (LLM). The resulting KGs, stored in a Neo4j database, capture both the internal topology of legal texts and their semantic nuances, enabling structured representations that support more meaningful analysis than unstructured raw text. To assess the effectiveness of this approach, comparative experiments were conducted on document similarity tasks, a core component of legislative workflows. Three scenarios were evaluated: (i) a text-only baseline using BERT-based sentence embeddings averaged across document sections; (ii) structure-aware KGs encoded via FastRP and GraphSAGE embeddings; and (iii) contextually enriched KGs also encoded via FastRP and GraphSAGE embeddings. Results using legislative proposals from the Legislative Assembly of Rio Grande do Norte (ALRN) show that while the text-based model achieved the highest precision, recall, and F1-scores, the KG-based representations provided interpretable, structure-driven insights. Contextual enrichment improved FastRP’s performance over structure-only graphs, while GraphSAGE performed best with structure-only representations, suggesting that LLM-derived relations may have introduced semantic noise for this graph model. Although the KGs were inherently heterogeneous, homogeneous graph algorithms were applied for simplicity, which may have limited performance. Nevertheless, the results demonstrate the feasibility of converting legislative documents to KGs, and the inclusion of structural information in embeddings was achieved, demonstrating potential for future improvements via heterogeneous models, advanced pooling strategies, or self-supervised pre-training. By bridging NLP and graph-based AI, this work advances approaches for legal document modeling, offering a pipeline for document similarity analysis and improving the legislative process efficiency.Doctoral Thesis Características de projeto ótimo de um sistema de geração híbrido baseado em máquina de indução com acionamento por meio de um Regulador Eletromagnético de Frequência(Universidade Federal do Rio Grande do Norte, 2025-10-17) Crisóstomo, Daniel Carlos de Carvalho; Medeiros Júnior, Manoel Firmino de; https://orcid.org/0000-0002-9408-3948; http://lattes.cnpq.br/4960078797028638; https://orcid.org/0000-0002-9752-688X; http://lattes.cnpq.br/5456416955432542; Medeiros, Arthur Salgado de; Pinheiro, Ricardo Ferreira; Ramos, Thales Augusto de Oliveira; Bezerra, Ubiratan HolandaThe need for the implementation of policies aimed at energy transition drives the renewable energy sector, prompting increased investments in this field. Globally and nationally, data reveal significant growth in the incorporation of wind and solar energy into power grids, fostering further technological advancements to ensure efficiency and quality in the generation process. This study optimizes a hybrid wind generation system using an Electromagnetic Frequency Regulator (EFR), aiming to minimize electrical losses, reduce component weight, and decrease reliance on secondary sources. The optimization explores the relationship between the number of poles in the generator and EFR, as well as the gearbox transmission ratio. The methodology combines numerical simulations, metaheuristic optimization (Ant Colony Optimization), electric machine design tools, and financial analysis. Results demonstrate that the optimal configuration heavily depends on the pole numbers and gearbox ratio, which directly affect the electrical parameters of the EFR and induction generator. Five topologies were analyzed, with the last proving the most efficient under the proposed criteria. The study provides a scalable methodology to assess the technical and economic feasibility of hybrid systems in diverse locations, contingent on wind data and generator specifications.Doctoral Thesis Deep Learning Applications in Geophysics: Data Matching and Inversion(Universidade Federal do Rio Grande do Norte, 2025-08-26) Perez, Katerine de Jesus Rincon; Souza, Samuel Xavier de; https://orcid.org/0000-0001-8747-4580; http://lattes.cnpq.br/9892239670106361; http://lattes.cnpq.br/0756224243207855; Carmo, Bruno Souza; https://orcid.org/0000-0002-2486-7026; http://lattes.cnpq.br/8129545750278083; Corso, Gilberto; https://orcid.org/0000-0003-1748-4040; http://lattes.cnpq.br/0274040885278760; Senger, Hermes; http://lattes.cnpq.br/3691742159298316; Araújo, João Medeiros de; https://orcid.org/0000-0001-8462-4280; http://lattes.cnpq.br/3061734732654188; Oliveira, Luiz Affonso Henderson Guedes de; https://orcid.org/0000-0003-2690-1563; http://lattes.cnpq.br/7987212907837941This work applies Deep Learning (DL) methodologies to solve geophysical problems, with a focus on Time lapse seismic, an essential technique for monitoring changes in reservoirs and supporting strategic management, using synthetic pre-salt seismic data acquired with Ocean Bottom Nodes (OBN). The first approach focused on mitigating one of the main limitations of TL seismic: data non-repeatability between surveys, caused by various environmental factors and equipment positioning variations, which can mask real reservoir changes. To address this challenge, we proposed Conditional Generative Adversarial Networks (cGANs). This methodology was developed to correct non-repeatability effects, improving data quality before applying Double Difference Full Waveform Inversion (DDFWI). As a second application, designed to estimate velocity variations and serve as a technique parallel to inversion, we constructed a synthetic dataset representing scenarios with different velocity anomalies. A Convolutional Neural Network (CNN) was trained using synthetic baseline and monitor seismograms generated from a pre-salt velocity model. The network received the time-lapse difference between baseline and monitor data as input, and produced the corresponding velocity model as output. To benchmark the approach, conventional Double-Difference Full Waveform Inversion (DDFWI) was also performed, providing a reference for computational cost, efficiency, and inversion quality. Results demonstrate that the CNN achieved accurate inversion performance while delivering significantly higher computational efficiency compared to the traditional DDFWI workflow. Both scenarios were trained with 80% of the data and tested with the remaining 20%. The predictions were evaluated using quantitative metrics such as the mean squared error (MSE), the normalized root mean squared error (NRMS), and the structural similarity index (SSIM), indicating a good inference capacity of the model.Doctoral Thesis K-Nearest Neighbors para detecção de anomalias e manutenção preditiva em sistemas de abastecimento hídrico(2025-08-27) Silva, João Pablo Santos da; Maitelli, André Laurindo; https://orcid.org/0000-0001-6083-7425; http://lattes.cnpq.br/0477027244297797; https://orcid.org/0000-0002-4173-1702; http://lattes.cnpq.br/7498666974011567; Dória Neto, Adrião Duarte; https://orcid.org/0000-0002-5445-7327; http://lattes.cnpq.br/1987295209521433; Salazar, Andrés Ortiz; https://orcid.org/0000-0001-5650-3668; http://lattes.cnpq.br/7865065553087432; Nascimento, João Maria Araújo do; https://orcid.org/0000-0002-4690-086X; http://lattes.cnpq.br/0572726574522312; Gabriel Filho, Oscar; http://lattes.cnpq.br/4171033998524192Este trabalho propõe a aplicação de metodologias de aprendizado de máquina e redes neurais com o objetivo de otimizar a operação de redes de abastecimento hídrico, com ênfase na detecção de anomalias hidráulicas, tais como vazamentos e falhas em sistemas de bombeamento. Diante da crescente necessidade de gestão eficiente de recursos hídricos, alinhada aos conceitos de cidades inteligentes, visando à sustentabilidade, redução de perdas e melhoria na qualidade da água, foram desenvolvidos modelos preditivos para identificação de anomalias e proposição de soluções que aumentem a eficiência operacional, reduzindo falhas e custos de manutenção, a partir da utilização dos dados de sensores instalados em uma rede de abastecimento. O método utilizado neste estudo empregou o algoritmo K-Nearest Neighbors (KNN) para análise correlacional e predição de falhas. A aplicação do algoritmo foi feita para prever a inoperância de bombas com base nos dados dos sensores e integrar diferentes abordagens para aprimorar a precisão dos resultados. O modelo priorizou a simplicidade, mantendo alta acurácia com o menor número possível de parâmetros de entrada. Os resultados demonstraram que a abordagem possibilita a detecção precoce de anomalias, a tomada de decisão rápida com base em dados em tempo real e a geração de percepções operacionais antes não identificadas, assim como a redução no índice de falhas operacionais, o que comprovou a eficácia do sistema. Conclui-se que a integração de aprendizagem de máquina e redes neurais otimiza a gestão de redes hídricas, assegurando abastecimento contínuo e qualidade da água. O modelo desenvolvido pode ser adaptado a diferentes contextos, reforçando sua relevância para aplicações industriais e gestores públicos.Doctoral Thesis Sistema vestível para reconhecimento de gestos em libras com base em superfície seletiva de frequência(Universidade Federal do Rio Grande do Norte, 2025-08-01) Oliveira, João Guilherme Domingos de; Silva Neto, Valdemir Praxedes da; D'Assunção, Adaildo Gomes; http://lattes.cnpq.br/4159638862269940; https://orcid.org/0000-0003-3867-3297; http://lattes.cnpq.br/4160231554601828; https://orcid.org/0000-0002-3092-7603; http://lattes.cnpq.br/6182090178676621; D'Assunção Júnior, Adaildo Gomes; https://orcid.org/0000-0002-6543-3187; http://lattes.cnpq.br/7359899329008024; Gomes Neto, Alfredo; https://orcid.org/0000-0001-5437-9093; http://lattes.cnpq.br/1403715441701958; Campos, Antonio Luiz Pereira de Siqueira; http://lattes.cnpq.br/1982228057731254; Silva, Gutembergue Soares da; http://lattes.cnpq.br/1214925346969928This work presents the design, prototyping, experimental validation, and analysis of a new wearable device based on the coupling of a textile frequency selective surface (FSS) to a data glove, developed for static hand gesture recognition applications. Initially, a novel textile FSS is proposed, employing 100% cotton fabric as the dielectric material and a commercial conductive thread, Shieldex 78f20. The performance of the FSS is evaluated using two computational methods: a full-wave method, implemented in commercial tools (ANSYS HFSS) based on the finite element method (FEM), and the approximate equivalent circuit method (ECM), in which a simplified equivalent circuit of the proposed geometry is presented. After computational and experimental validation, the FSS was coupled to a data glove and applied to the recognition of different static hand gestures. As reported in the literature, several methods can be employed for this purpose, involving parameters related to variations in data obtained through optical, radiofrequency (RF), or computer vision signal measurements. In this work, the proposed method obtains data through an RF setup. Initially, five static gestures were measured, and based on the collected data, the variation in the magnitude of the S21 transmission coefficient was adopted as the classification metric. To numerically validate the results, a statistical analysis was performed using the t-Student test. The results indicated that 70% of the measured gestures presented significant differences among each other, considering a p-value < 0.005, highlighting the efficiency of the proposed device for this application. Additionally, a robustness analysis was conducted with two Brazilian Sign Language (Libras) gestures with small lexical differences, the letters “a” and “s.” The experimental results and subsequent statistical analysis showed significant differences when compared. The experimental values were compared to the simulated results, showing good agreement. Therefore, the findings confirm the potential of the proposed approach for applications in gesture recognition, human–machine interaction, virtual gaming interfaces, and augmented reality.Doctoral Thesis Implementação de técnica Active Disturbance Rejection Control (ADRC) no controle de velocidade do Regulador Eletromagnético de Frequência (REF)(Universidade Federal do Rio Grande do Norte, 2025-07-02) Oliveira, Gabriel Pereira de; Salazar, Andrés Ortiz; https://orcid.org/0000-0001-5650-3668; http://lattes.cnpq.br/7865065553087432; http://lattes.cnpq.br/2340427873121127; Maitelli, André Laurindo; https://orcid.org/0000-0001-6083-7425; http://lattes.cnpq.br/0477027244297797; Villarreal, Elmer Rolando Llanos; https://orcid.org/0000-0002-3059-3340; http://lattes.cnpq.br/5447452711376721; Paiva, José Alvaro de; https://orcid.org/0000-0002-9148-6178; http://lattes.cnpq.br/6136888701626547; Ferreira, Jossana Maria de Souza; http://lattes.cnpq.br/7691693293898376; Pinheiro, Ricardo Ferreira; http://lattes.cnpq.br/4890839733220743The continuous increase in energy demand, combined with growing social and political concerns about sustainability, has driven research and development in Renewable Energy Sources (RES), with particular emphasis on wind and photovoltaic (PV) generation. The temporal complementarity of different sources and the possibility of utilizing existing infrastructure favor the adoption of hybrid generation systems. In this context, the Electromagnetic Frequency Regulator (EFR) emerges as a promising alternative, enabling electromechanical hybridization through a modified induction machine whose rotating armature is coupled to the wind turbine, while a second primary energy source supplies electrical power to the machine via a Voltage Source Inverter (VSI). This physical arrangement provides the EFR with important advantages over conventional architectures based on static converters, such as greater electromechanical inertia and reduced harmonic injection into the power grid. Additionally, the EFR simplifies the design of the gearbox in wind systems, resulting in a more compact and efficient setup. On the other hand, controlling the EFR is challenging due to the strong presence of external disturbances, such as variations in wind profile and the behavior of connected loads. Since the EFR supplies the grid or local loads through a synchronous generator, it becomes essential to maintain a constant rotor speed in order to ensure stability and power quality. In this work, a control strategy based on Field-Oriented Control (FOC) is proposed, in which the outer speed control loop replaces the traditional PI controller—commonly used in machine drives—with two more robust alternatives: Active Disturbance Rejection Control (ADRC) and its linear version (LADRC). These controllers employ an Extended State Observer (ESO) to estimate and reject, in real time, the internal and external disturbances affecting the system, thereby simplifying the plant dynamics for the controller design. The effectiveness of the proposed technique is validated through computational simulations using the MATLAB/Simulink environment and experimental bench tests, considering different operating conditions and disturbance scenarios.Doctoral Thesis Estratégia Grid-Forming para geração eólica em microrredes utilizando regulador eletromagnético de frequência(Universidade Federal do Rio Grande do Norte, 2025-06-30) Dantas Neto, José Raimundo; Salazar, Andrés Ortiz; https://orcid.org/0000-0001-5650-3668; http://lattes.cnpq.br/7865065553087432; http://lattes.cnpq.br/0632687844877314; Maitelli, André Laurindo; https://orcid.org/0000-0001-6083-7425; http://lattes.cnpq.br/0477027244297797; Fonseca, Diego Antonio de Moura; Larrea, José Andres Santisteban; Pinheiro, Ricardo Ferreira; Araújo, Valbério Gonzaga deThe increasing use of renewable energy sources (RES), especially solar and wind, has driven significant transformations in the Electric Power System (EPS), which is now organized in a more distributed manner through microgrids. This new paradigm requires control solutions that ensure stability, power quality, and robustness in operation. Traditionally, distributed generation (DG) uses grid-following control strategies, which the RES connects to the grid following its frequency and voltage. However, this approach contributes to the EPS instability in scenarios with high RES penetration. In contrast, grid-forming strategies have stood out for allowing the DG itself to define the voltage and frequency parameters, promoting greater stability, dynamic support, and resilience. Despite the advances, the use of power converters in the interface with the grid presents considerable challenges, such as low inertia and dynamic response limitations. Many control approaches aim to emulate the behavior of a synchronous generator (SG) to mitigate these problems. In this context, the use of an Electromagnetic Frequency Regulator (EFR) emerges as a promising solution for wind power systems (WPS), as it enables the use of an SG in the connection to the microgrid. This configuration eliminates the interface converter, promoting a robust electromechanical coupling, capable of providing inertia to the system, contributing to its stability and power quality. Thus, this thesis proposes an interconnection strategy for a WPS using an EFR/SG set operating in grid-forming mode. The strategy provides support to the main grid when connected, through the power flow regulation using the droop control technique. In addition, it ensures frequency and voltage inertial response, especially during islanding conditions. The start-up and synchronization process is also addressed, aligning the DG and grid voltage vectors. The validation of the proposed strategy is carried out through simulations in the Matlab/Simulink platform and experimental tests conducted at the Petroleum Assessment and Measurement Laboratory (LAMP), demonstrating the effectiveness and robustness of the presented solution.Doctoral Thesis High-performance graphene-based devices for terahertz and optical communications(Universidade Federal do Rio Grande do Norte, 2025-07-18) Souto, Gabriel Amaral; Silva, José Patrocinio da; https://orcid.org/0000-0003-1843-7879; http://lattes.cnpq.br/5753289728835624; https://orcid.org/0000-0001-7674-4418; http://lattes.cnpq.br/3165936278609508; Campos, Antonio Luiz Pereira de Siqueira; http://lattes.cnpq.br/1982228057731254; Queiroz Júnior, Idalmir de Souza; Mendonça, Laercio Martins de; Holanda, Samanta Mesquita deThe continuous evolution of information technology demands increasingly innovative, efficient, and reliable solutions for data processing and transmission. With the exponential growth in high-speed communication requirements and large-scale data transfer, it becomes essential to explore technological alternatives capable of operating at higher frequency bands. The transition to elevated frequency ranges—such as the upper gigahertz (GHz) and terahertz (THz) bands—enables the development of high-performance devices capable of supporting significantly higher data rates with reduced latency. In this context, graphene has emerged as one of the most promising materials for high-frequency applications. Unlike conventional noble metals, graphene offers several advantageous properties: it exhibits substantially lower losses, its atomic structure provides high mechanical flexibility and robustness, and its electrical conductivity can be dynamically tuned. Furthermore, graphene possesses remarkable resistance to oxidation and moisture, making it suitable for operation in harsh environments. These attributes make it particularly attractive for devices operating in the GHz–THz range, where precision and efficiency are critical. This work investigates two innovative high-performance graphene-based devices developed for quantum cascade laser (QCL) systems operating in the terahertz range. The first is a cardioid-shaped microstrip antenna designed to maximize directivity and transmission efficiency at THz frequencies, employing a multilayer graphene structure to enhance electromagnetic wave propagation. The second is a graphene-based rib waveguide, designed as a plasmonic biosensor with high sensitivity to refractive index variations, making it particularly effective for chemical and biological detection through the excitation of strongly confined surface plasmons. Graphene’s dynamic conductivity was obtained using the Kubo formalism and subsequently modeled with a Drude-like approximation to define final parameters. The antenna employed a cardioid-shaped radiating element on a silica substrate, designed using the Full-Wave Method (FWM) for initial sizing, followed by Finite Element Method (FEM) simulations to analyze substrate height effects. The plasmonic sensor consisted of a rib waveguide with a tantalum pentoxide (Ta2O5) substrate and a niobium pentoxide (Nb2O5) ridge, overlaid with multilayer graphene. This structure was designed to detect dimethyl sulfoxide (DMSO) concentrations from 50% to 100% in 10% increments. Localized surface plasmon resonance (LSPR) occurred at the graphene–analyte interface, induced by free-electron oscillations in graphene coupled to electromagnetic fields. For DMSO detection, the sensor was tuned to operate at a wavelength of 0.6329 µm. Publication data analysis shows that research in terahertz technology is concentrated between 2013 and 2025, with the earliest record in 1991 marking the beginning of the dataset. Steady growth is observed, with notable peaks in 2016, 2021, and 2024—the latter recording the highest number of publications. The intermediate years, such as 2018 and 2019, maintained regular research activity, reflecting sustained interest and maturity in THz device applications. At λ = 0.6329 µm, the electric field magnitude increased with DMSO concentration, from 4.69 V/m (50%) to 7.96 V/m (pure DMSO), indicating enhanced field confinement and plasmonic response at higher concentrations. Confinement loss (CL) also rose with concentration, ranging from 21.62 dB/cm to 22.80 dB/cm. Total spectral sensitivity remained consistent at approximately 2222 nm/RIU, with maximum and minimum values of 2224.37 nm/RIU (pure DMSO) and 2221.07 nm/RIU (50%), respectively. Average spectral sensitivity increased from 20283.740 nm/RIU (50%) to 20314.470 nm/RIU (100%), while local sensitivity ranged from 22039.610 to 22071.460 nm/RIU. The figure of merit (FOM) remained high, averaging 2050 RIU−1, with the highest value of 2478 RIU−1 observed for 50% DMSO. Substrate thickness strongly influenced antenna resonant frequency. At 25.0 µm, the resonance was 3.73 THz with 73 GHz bandwidth and −18.30 dB reflection. For 30 µm, it was 1.653 THz with 133 GHz bandwidth and −30.11 dB reflection. At 37.5 µm (thinnest standard commercially available), resonance occurred at 2.385 THz with 21 GHz bandwidth and −43.00 dB reflection. At 40 µm, resonance was 2.287 THz with 9 GHz bandwidth and −15.12 dB reflection. The structure allowed thickness adjustments in 5 µm increments, enabling spectral response tuning. At 2.385 THz, the antenna exhibited 1.967 dB gain, 970 mW radiated power, and a 1.970 front-to-back ratio. The radiation pattern displayed a main lobe along the z-axis, characteristic of slot or Vivaldi antennas, though high-frequency beam behavior warrants further investigation due to its nonlinear nature. The multilayer graphene antenna demonstrated efficient operation in the far-infrared (FIR) range, as indicated by simulation-derived performance parameters. The reflection coefficient (S11) remained below −10 dB at the target frequency, confirming good impedance matching and minimal signal reflection. Gain and radiated power results indicated suitable directivity, crucial for high-frequency optical communication systems. The electric field distribution on the patch revealed a well-defined resonant mode, characteristic of efficient energy-coupling structures. For the photonic waveguide, operation in the same spectral range was confirmed by FEM analysis of plasmonic modes. The electric field was strongly confined at the graphene–dielectric interface, with confinement losses within acceptable limits for practical operation. Effective refractive index analysis confirmed propagation feasibility for different DMSO concentrations, reinforcing the device’s suitability for optical biosensing and compatibility with graphene-based emission sources. Antenna–waveguide compatibility was assessed through spectral and spatial mode matching. Dynamic tuning of graphene’s chemical potential via electrostatic gating or chemical doping enabled alignment of antenna emission peaks with supported waveguide modes. This tunability ensured efficient coupling, minimized insertion losses, and facilitated integration into QCL-based photonic circuits. The results confirm both independent device performance and their functional integration into compact, reconfigurable infrared systems.Doctoral Thesis Uma arquitetura baseada em aprendizado profundo com mecanismos de atenção para a tradução contínua da língua brasileira de sinais em contextos sem intérpretes(Universidade Federal do Rio Grande do Norte, 2024-12-20) Silva, Diego Ramon Bezerra da; Gonçalves, Luiz Marcos Garcia; https://orcid.org/0000-0002-7735-5630; http://lattes.cnpq.br/1562357566810393; https://orcid.org/0000-0002-1037-9953; http://lattes.cnpq.br/1318177376490124; Silva, Bruno Marques Ferreira da; http://lattes.cnpq.br/7878437620254155; Souza Filho, Guido Lemos de; https://orcid.org/0000-0001-5834-5237; http://lattes.cnpq.br/6614550860293610; Araújo, Tiago Maritan Ugulino de; https://orcid.org/0000-0002-5953-5435; http://lattes.cnpq.br/6347743344931103; Campos, Virginia Pinto; http://lattes.cnpq.br/1962361122489165In Brazil, the deaf represent about 5% of the population approximately 9.7 million Brazilians. Despite the Brazilian Sign Language (Libras) being recognized as an official language in Brazil, the knowledge and mastery of Libras among the non-deaf is an obstacle, which ends up creating language barriers in accessing basic rights, especially in accessing health services. This has motivated the development of government policies that oblige service providers to provide Libras interpreters to enable access to these services by the deaf community. However, human-based approaches have a high implementation and maintenance cost. From this perspective, it is necessary to develop research and automated methodologies for automatic translation of Libras. Thus in this work, we proposed a methodology for continuous translation of Libras. The proposed methodology does not require any additional hardware, relying entirely on images or image sequences (videos). In addition, a new dataset for Continuous Sign Language Recognition (CSLR) was introduced, containing 10500 videos of 105 distinct sentences in the context of clinical screening. The evaluation experiments achieve a WER of 21.62, while maximum accuracy of about 92.68% for a set of tests with videos and signers never seen by the model during training.Doctoral Thesis Desenvolvimento de ferramenta computacional para biossensores baseados em ressonância de plasmon de superfície utilizando métodos híbridos e aprendizado de máquina(Universidade Federal do Rio Grande do Norte, 2025-06-26) Villarim, Mariana Rodrigues; Belfort, Diomadson Rodrigues; Ando Júnior, Oswaldo Hideo; https://orcid.org/0000-0002-6951-0063; http://lattes.cnpq.br/3515465412634126; https://orcid.org/0000-0001-5722-7394; http://lattes.cnpq.br/2241184480091242; http://lattes.cnpq.br/1234708675762453; Maciel, Joylan Nunes; https://orcid.org/0000-0003-0725-6917; http://lattes.cnpq.br/1177414528561833; Cavallari, Marco Roberto; https://orcid.org/0000-0002-1345-754X; http://lattes.cnpq.br/9041429608835546; Carmo, João Paulo Pereira do; https://orcid.org/0000-0001-7955-7503; http://lattes.cnpq.br/5589969124054528; Catunda, Sebastian Yuri Cavalcanti; http://lattes.cnpq.br/0873496251879638Doctoral Thesis Integrating textual queries with aI-based object detection: a compositional prompt-guided approach(2025-06-13) Silva Júnior, Silvan Ferreira da; Martins, Allan de Medeiros; Silva, Ivanovitch Medeiros Dantas da; https://orcid.org/0000-0002-0116-6489; http://lattes.cnpq.br/3608440944832201; https://orcid.org/0000-0002-9486-4509; http://lattes.cnpq.br/4402694969508077; http://lattes.cnpq.br/0669123874358910; Guimarães, João Paulo Ferreira; http://lattes.cnpq.br/8309552415999065; Villanueva, Juan Moisés Mauricio; https://orcid.org/0000-0002-8760-9390; http://lattes.cnpq.br/1446817462218646; Oliveira, Luiz Affonso Henderson Guedes de; https://orcid.org/0000-0003-2690-1563; http://lattes.cnpq.br/7987212907837941In the field of computer vision, object detection and recognition play a central role in many applications that support automated decision-making. Over recent years, new algorithms and methodologies have emerged to further enhance the automatic identification of target objects. In particular, the rise of deep learning and language models has opened many possibilities in this area, although challenges in contextual query analysis and human interactions persist. This thesis presents a novel neuro-symbolic object detection framework that aligns object proposals with textual prompts using a deep learning module while enabling logical reasoning through a symbolic module. By integrating deep learning with symbolic reasoning, object detection and scene understanding are considerably enhanced, enabling complex, query-driven interactions. Using a synthetic 3D image dataset, the results demonstrate that this framework effectively generalizes to complex queries, combining simple attribute-based descriptions without explicit training on compound prompts. We present the numerical results and comprehensive discussions, highlighting the potential of our approach for emerging smart applications.Doctoral Thesis Detecção e localização de crises epilépticas em sinais de EEG utilizando aprendizado de máquina e inteligência artificial explicável(Universidade Federal do Rio Grande do Norte, 2025-05-30) Vieira, Jusciaane Chacon; Oliveira, Luiz Affonso Henderson Guedes de; https://orcid.org/0000-0003-2690-1563; http://lattes.cnpq.br/7987212907837941; Teixeira, César; Gendriz, Ignacio Sanchez; http://lattes.cnpq.br/6338710569530857; Silva, Ivanovitch Medeiros Dantas da; https://orcid.org/0000-0002-0116-6489; http://lattes.cnpq.br/3608440944832201; Fernandes, Marcelo Augusto Costa; https://orcid.org/0000-0001-7536-2506; http://lattes.cnpq.br/3475337353676349Epilepsy is a neurological condition that affects millions of people worldwide and significantly impacts individuals’ quality of life. Epileptic seizures, which are transient events, vary in manifestations, including motor, sensory, and consciousness alterations, and represent a challenge both in diagnosis and management. This work proposes an innovative methodology for the detection and localization of epileptic seizures, using machine learning and explainable artificial intelligence approaches to optimize the identification process. The proposal is divided into two approaches: a generalist one, which uses simplified models with explainable feature and channel reduction, and a specific one, which personalizes detection for each patient based on a single electroencephalogram (EEG) channel. In the generalist approach, feature and channel reduction was explored, achieving performance above 0.95 in accuracy, precision, recall, and f1-score metrics using only six features and five channels. The use of the SHAP method allowed for the interpretation of each feature’s contribution by channel, reinforcing the explainability of the models and aligning the results with the visual knowledge of EEG specialists. The methodology proved effective, ensuring good generalization to the dataset with different patients. In the specific approach, a bipolar montage with adjacent electrodes was introduced, creating 58 channel combinations, and a centrality measure with topographic visualization was applied to identify the most relevant channels for each patient. Additionally, a 3-second temporal filter was developed to reduce the model’s false positives. A personalized supervised learning model, extreme gradient boosting, was trained for each patient using only one EEG channel. The results for the three investigated patients were remarkable, with accuracy scores of 1; 0.99; and 0.88, highlighting the feasibility of seizure detection using a single channel, considering the topographic location of seizures in each individual. This study highlights the potential to substantially reduce the number of features and channels required for epileptic seizure detection without compromising accuracy, and reinforces the importance of personalized models for each patient. Furthermore, the research contributes to the advancement of wearable devices for continuous epilepsy monitoring, facilitating the detection and localization of patients with epilepsy.Doctoral Thesis Multi-agent reinforcement learning for inter-cell interference management in hotspots scenarios(2025-05-05) Rêgo, Iago Diógenes do; Sousa Júnior, Vicente Angelo de; Conceição, Dario Vieira; https://orcid.org/0000-0001-8418-6778; http://lattes.cnpq.br/1142876099153789; https://orcid.org/0000-0003-2859-6136; http://lattes.cnpq.br/6358312955522220; http://lattes.cnpq.br/7295912447680309; Cavalcante, André Mendes; http://lattes.cnpq.br/3939185965976929; Campos, Antonio Luiz Pereira de Siqueira; http://lattes.cnpq.br/1982228057731254; Venâncio Neto, Augusto José; http://lattes.cnpq.br/1467664612924239; Silva, Yuri Carvalho Barbosa; https://orcid.org/0000-0002-1795-6004; http://lattes.cnpq.br/7399179037419892Inter-cell interference (ICI) remains a critical challenge in mobile networks. Although current and future standards are rapidly evolving, the constant increase in data demand, the emergence of new use cases, the coexistence of multiple technologies, and the dynamic aspect of urban environments intensifies the impact of interference on system performance. ICI becomes especially challenging in dense deployments and in scenarios with zones of high user densities, referred to as hotspots. Fractional Frequency Reuse (FFR) is a well-established technique to mitigate ICI in OFDMA-based networks such as LTE and 5G, but traditional static configurations often fail to adapt to dynamic interference patterns. This thesis proposes a dynamic interference coordination framework based on reinforcement learning, designed to enhance the adaptability and performance of FFR techniques. The solution consists of a hierarchical multi-agent architecture, where two reinforcement learning agents operate in coordination, without direct exchange of information, to jointly control the allocation of bandwidth and user classification via the RSRQ threshold. The proposed approach was evaluated through network simulations using ns-3, across two different scenarios representing dense urban environments and massive connection conditions. The results show that the proposed framework consistently outperforms static and single-agent baselines, achieving throughput gains of up to 99.4%, particularly under high-interference conditions and for low-performing users. Furthermore, its modular design allows integration of different learning strategies. While Q-Learning agents delivered the highest performance, Multi-Armed Bandit (MAB) agents achieved comparable results with significantly lower computational complexity. By combining classical ICIC techniques with reinforcement learning, this work presents a flexible and low-overhead interference mitigation strategy that can adapt to diverse and evolving network conditions.Doctoral Thesis Novas configurações de sensores planares para a caracterização não-destrutiva de materiais magneto-dielétricos em micro-ondas(Universidade Federal do Rio Grande do Norte, 2025-01-30) Nogueira, Jurgen Klinsmann Azevedo; D'Assunção, Adaildo Gomes; Silva Neto, Valdemir Praxedes da; https://orcid.org/0000-0003-3867-3297; http://lattes.cnpq.br/4160231554601828; http://lattes.cnpq.br/4159638862269940; http://lattes.cnpq.br/3159722016018256; Costa, Cláudio Pereira da; https://orcid.org/0000-0003-2073-8291; http://lattes.cnpq.br/7571127714246653; Carvalho, Joabson Nogueira de; https://orcid.org/0000-0001-8173-1160; http://lattes.cnpq.br/3511078820928127; Oliveira, João Bosco Lucena de; https://orcid.org/0000-0002-8946-2521; http://lattes.cnpq.br/1360004695124740; Mendonça, Laércio Martins de; https://orcid.org/0000-0001-7861-3761; http://lattes.cnpq.br/1853488415531363This work presents the design and fabrication of two new configurations of planar microstrip devices for sensing applications and to characterize non-destructively the complex permittivity of magneto-dielectric materials. Planar microwave sensors have demonstrated important functionalities in several applications, such as in the food, industrial and biomedical sectors. Microwave sensors have the advantages of low cost and ease of fabrication, in addition to the possibility of integration with other devices. The first proposed sensor is based on a two-port planar microstrip device with the insertion of a Complementary Split Ring Resonator (CSRR) in its patch, this element is responsible for determining the electrical properties of the materials under test (MUT). Two samples of materials widely known in the literature were investigated. With the first proposed sensor, it was possible to obtain the complex permittivity of two solid materials. The second proposed sensor is based on a bioinspired microstrip planar antenna with a change in the ground plane structure (Defected Ground Structure – DGS), in order to obtain, also in a non-destructive way, the magneto-dielectric characteristics, i.e., the electrical permittivity and magnetic permeability, of nine solid materials also known in the literature, using the same resonator for both characterizations. The proposed sensor has a patch bioinspired by the Spondias Tuberosa leaf, and a CSRR was inserted in the center of the ground plane, responsible for the permittivity characterization. The permeability was obtained by means of slits positioned close to the CSRR. In this work, the sensitivities of the two proposed sensors were also analyzed, presenting excellent results. The sensors were analyzed by means of simulations in the commercial software Ansys HFSS. The prototypes were manufactured using the FR-4 dielectric, which has a relative electrical permittivity of 4.4, a loss constant of 0.02 and a thickness of 1.57 mm, and the measurements were performed using the E5071C vector network analyzer (VNA) on nine different samples (MUT) of dielectric and magnetic materials. The simulated and measured results were discussed and compared with values found in the literature, showing good agreement.Doctoral Thesis A configurable dependability library for high-performance computing iterative applications with interruption detection, data preservation and failover capabilities(Universidade Federal do Rio Grande do Norte, 2024-10-04) Santana, Carla dos Santos; Souza, Samuel Xavier de; https://orcid.org/0000-0001-8747-4580; http://lattes.cnpq.br/9892239670106361; https://orcid.org/0000-0003-3328-0056; http://lattes.cnpq.br/4697610292983660; Bianchini, Calebe de Paula; Tadonki, Claude; Chauris, Herve; Taufer, Michela; Navaux, Philippe Olivier Alexandre; Barros, Tiago Tavares LeiteHigh-performance computing, a dynamic field within computer science, provides the processing power necessary for algorithms across diverse domains. Large-scale supercomputers are indispensable for tackling complex problems; however, their size and complexity make them susceptible to failure. This underscores the criticality of employing fault tolerance techniques to mitigate the impact of interruptions or failures. These methods are instrumental in addressing hardware and software malfunctions and preemptive scenarios. Given the imperative for fault tolerance, we present new methodologies for improving fault tolerance in bulk synchronous programs. These new methodologies are presented as the Dependability Library for Iterative Applications. This library offers a versatile solution that combines data conservation at the application level, fault detection, and failover capabilities. The proposed library simplifies the integration of fault tolerance abilities into the applications, offering high configurability options. This thesis presents data conservation methodologies, including application-level checkpointing and process data replication, to ensure reliability by allowing a backup unit to take over in case of failure. This work also presents fault detection methods such as termination signal detection and heartbeat monitoring with inexpensive communication to trigger the data conservation only if there is a possibility of failure; this approach permits low overhead. The proposed library is compatible with user-level failure mitigation, which allows failover capabilities; in other words, the programs can continue operating after crashes, minimizing downtime and ensuring continuous operation. Our proposal was successfully applied to the geophysical problem of full-waveform inversion, a standard algorithm for oil and gas exploration geophysics processing. This application serves as a high-performance practical scenario for analysis, demonstrating the real-world applicability of the library. All methods were rigorously validated, and the overhead in this problem was thoroughly analyzed using more realistic examples. In our experiments, the application did not lose all data processed until the failure moment, and it could continue execution even in the presence of node failure, with minimal overhead. This work also shows other case studies in the initial stage of applying the library and discusses some fault tolerance concepts and related works.Doctoral Thesis Transmissão vertical da sífilis no Brasil: análises baseadas em ciência de dados na saúde pública(Universidade Federal do Rio Grande do Norte, 2024-10-30) Silva, Rodrigo Dantas da; Valentim, Ricardo Alexsandro de Medeiros; https://orcid.org/0000-0002-9216-8593; http://lattes.cnpq.br/3181772060208133; https://orcid.org/0000-0002-2549-2414; http://lattes.cnpq.br/1947688093671056; Campos, Antonio Luiz Pereira de Siqueira; http://lattes.cnpq.br/1982228057731254; Coutinho, Karilany Dantas; Santos, Marquiony Marques dos; Morais, Antonio Higor Freire de; Lima, Thaisa Gois Farias de Moura Santos; Valentim, Janaína Luana Rodrigues da Silva; Santos, João Paulo Queiroz Dos; 01065547439This study examines the shortcomings of the Clinical Protocol and Therapeutic Guidelines (PCDT) for the management of gestational and congenital syphilis in Brazil, demonstrating that, in addition to the limitations of the protocol itself, there are also serious structural failures in the healthcare system. The research aims to prove that the inefficacy of the PCDT, combined with a lack of integration between levels of care and health surveillance, contributes to the increase in vertical transmission rates of syphilis. Using Stochastic Petri Nets (SPN) modeling and data from SINAN, treatment and diagnostic scenarios were simulated, revealing critical bottlenecks such as inadequate treatment for pregnant women and insufficient case monitoring. The study highlights that the failure to apply the guidelines, alongside poor prenatal care and a lack of proper training for healthcare professionals, is one of the greatest challenges in controlling congenital syphilis in the country.Doctoral Thesis A novel deep neural network technique for drug-target interaction prediction(Universidade Federal do Rio Grande do Norte, 2024-12-10) Souza, Jackson Gomes de; Fernandes, Marcelo Augusto Costa; Barbosa, Raquel de Melo; https://orcid.org/0000-0001-7536-2506; http://lattes.cnpq.br/3475337353676349; https://orcid.org/0000-0003-0665-7153; http://lattes.cnpq.br/7022849614714429; Villén, Fátima García; Silva, Lucileide Medeiros Dantas da; Coutinho, Maria Gracielly FernandesDrug discovery (DD) is a time-consuming and expensive process. Thus, the industry employs strategies such as drug repositioning and drug repurposing, which allows the application of already approved drugs to treat a different disease, as occurred in the first months of 2020, during the COVID-19 pandemic. Prediction of drug-target interaction is an essential part of the DD process because it can accelerate it and reduce required costs. DTI prediction performed in silico have used approaches based on molecular docking simulation, similarity-based and network and graph based. This paper presents MPS2ITDTI, a DTI prediction model obtained from research conducted in the following steps: the definition of a new method for representing/encoding molecule and protein sequences into images; and the definition of a deep-learning approach based on a convolutional neuralnetwork in order to create a new method for DTI prediction. The results of this research indicate that the image-based representation of molecule and protein sequences is a viable alternative to the NLP-based approaches and, as such, does not adopt an embedding layer in the neural network. The training results conducted with the Davis and KIBA datasets show that MPS2IT-DTI is viable compared to other state-of-the-art (SOTA) approaches in terms of performance and complexity of the neural network model. Regarding the Davis dataset, the results of the experiments indicate a concordance index (CI) of 0.876 and a MSE of 0.276; with the KIBA dataset, 0.836 and 0.226, respectively. Finally, the experimental results utilizing the BindingDB dataset and six core proteins of SARS-CoV-2 suggest that MPS2IT-DTI performs comparably with state-of-the-art methodologies for the repurposing of clinically approved antiviral agents in the context of COVID-19 treatment.Doctoral Thesis Reconfigurable hardware architecture for SHA-256 hashing in blockchain and IoT applications(Universidade Federal do Rio Grande do Norte, 2024-12-11) Santos Júnior, Carlos Eduardo de Barros; Fernandes, Marcelo Augusto Costa; Silva, Sérgio Natan; https://orcid.org/0000-0001-7536-2506; http://lattes.cnpq.br/3475337353676349; http://lattes.cnpq.br/1334493042199015; Dias, Leonardo Alves; Silva, Lucileide Medeiros Dantas da; Coutinho, Maria Gracielly FernandesAs IoT device usage continues to expand, ensuring secure, low-latency data exchange has become essential, driving research into blockchain-based solutions to meet these requirements. Addressing this demand, this thesis presents a reconfigurable hardware architecture for the SHA-256 hash algorithm, focusing on blockchain and IoT applications, utilizing Field Programmable Gate Arrays (FPGAs) as the target hardware to maximize performance and efficiency in data security processes. The proposed FPGA implementation provides adaptability across various environments, from network servers to energyconstrained IoT devices. Key innovations in this proposal include a multicore parallelism system that optimizes the use of available FPGA resources and a structured analysis of resource consumption, considering both clock frequency and throughput. Additionally, the thesis provides a power consumption analysis, comparing power efficiency across different hardware architectures. The proposed design achieved the implementation of 16 parallel cores on a Xilinx Virtex 6 xc6vlx240t-1ff1156 FPGA, reaching a maximum throughput of 1.4Gbps and dynamic power consumption of 0.452W. This performance represents up to 16x speedup over previous FPGA models and a reduction of up to 234.52x in dynamic power consumption compared to implementations from prior research. Additional comparisons were conducted with other hardware architectures, such as 8- and 16-bit microcontrollers, general-purpose processors, and GPUs. The results underscore the versatility and scalability of FPGA-based SHA-256 implementations for applications requiring high throughput and power efficiency, establishing this work as a significant contribution to information security and computational performance in IoT environments within a blockchain context.