ESTiG - Publicações em Proceedings Indexadas à WoS/Scopus
Permanent URI for this collection
Browse
Recent Submissions
- Analysis of DC-DC converters for integrated photovoltaic solar-powered electric bus charging systemsPublication . Soares, Orlando; Forigo, Paula W.This article presents a theoretical study and validation of the feasibility of integrating photovoltaic (PV) systems to charge electric vehicles, with a focus on an electric bus as the model for analysis. The work further explores the design of an energy conversion system employing DC-DC converters, analyzing various converter types and configurations to maximize energy transfer efficiency. The study identifies optimal converter topologies and control strategies to enhance system performance. Finally, the proposed models are validated through simulations in Matlab/Simulink, incorporating lithium-ion batteries as the energy storage system. The article aims to demonstrate how integrating photovoltaic energy into electric vehicle charging can serve as a sustainable and environmentally friendly solution, offering practical insights for improving energy efficiency in public transportation systems.
- Design and development of a differential drive platform for dragster competitionPublication . Grilo, Vinicius; Ferreira, Edilson; Barbosa, Ana; Chellal, Arezki Abderrahim; José L. LimaRobotics competitions have been increasing in the last years since they bring several impacts on students education, such as technical skill development, teamwork, resilience and decision making withing the STEM skills. The article highlights the significance of robotics competitions as platforms for fostering innovation and driving advancements in the field of robotics. This article primarily focuses on the development of a robot in the Dragster category for the 2023 Portuguese Robotics Open. It outlines the strategies devised to tackle the competition’s challenges and discusses the obstacles encountered along with the corresponding solutions employed. The article delves into the specific details of the challenges faced and the iterative processes undertaken to enhance the robot’s performance and functionalities. By sharing the insights gained from the project, future proposals for iterations of the robot will be presented, aiming to further augment its features and overall performance while sharing knowledge with other teams and community.
- Nonlinear control of mecanum-wheeled robots applying h∞ controllerPublication . Chellal, Arezki Abderrahim; Braun, João A.; Lima, José; Gonçalves, José; Valente, António; Costa, PauloMecanum wheeled mobile robots have become relevant due to their excellent maneuverability, enabling omnidirectional motion in constrained environments as a requirement in industrial automation, logistics, and service robotics. This paper addresses a low-level controller based on the H-Infinity (H∞) control method for a four-wheel Mecanum mobile robot. The proposed controller ensures stability and performance despite model uncertainties and external disturbances. The dynamic model of the robot was developed and introduced in MATLAB to generate the controller. Further, the controller’s performance is validated and compared to a traditional PID controller using the SimTwo simulator, a realistic physics-based simulator with dynamics of rigid bodies incorporating non-linearities such as motor dynamics and friction effects. The preliminary simulation results show that the H∞ reached a time-independent Euclidean error of 0.0091 m, compared to 0.0154 m error for the PID in trajectory tracking. Demonstrating that the H∞ controller handles nonlinear dynamics and disturbances, ensuring precise trajectory tracking and improved system performance. This research validates the proposed approach for advanced control of Mecanum wheeled robots.
- A GPU implementation of the analog ensemble methodPublication . Crico, Ruben; Charles, Ines; Balsa, Carlos; Rufino, JoséThe Analog Ensemble (AnEn) method can be used to reconstruct incomplete time series using correlated series. Since the AnEn method may use data including several variables through long periods of time, its storage and computational cost may be substantial, slowing down reconstructions. This paper presents a full GPU implementation of the AnEn method, based on PyCUDA, that leads to a significant a cceleration of its execution. The implementation resorts to several techniques that seek to minimize the consumption of GPU global memory in the various steps of the AnEn algorithm, thus making room for larger input datasets. This is further reinforced by the use of batch processing as a way to automatically adapt the datasets size to the GPU memory available. The GPU implementation was tested on a meteorological dataset spanning 10 years, exhibiting a 30-fold speedup in the reconstruction time against a comparable CPU-based multicore version executed with up to 48 cores. The impact on the reconstruction error of changes on several important parameters of the implementation was also assessed.
- Categorizing Students of the MathE Platform: A Fuzzy Clustering PerspectivePublication . Leite, Gabriel A.; Azevedo, Beatriz Flamia; Pacheco, Maria F.; Fernandes, Florbela P.; Pereira, Ana I.Active learning and technology integration offer enhanced student engagement and adaptive learning, accommodating diverse preferences. This work uses fuzzy clustering method to analyze the data of students who answer questions on the MathE platform. To do this, the Fuzzy c-means algorithm was used, which allows flexibility and adaptability in the clustering partitioning, especially in situations where data elements may exhibit overlapping characteristics or belong to multiple categories. Thereby, two datasets are considered: the first is composed of 121 students who answered questions from the Vector Space subtopic, and the second dataset comprises the answers of 297 students who answered to any topic or subtopic of the platform. The results show that the fuzzy clustering method is appropriate for analyzing the student’s data since most students are highly associated with more than one cluster. Besides, the findings can support the formulation of intervention strategies to improve the student’s academic achievement.
- A Neural Network-Based Approach to Identifying Wrinkles and Recommending Cosmetic ProductsPublication . Tonello, Guilherme de M.; Alves, Paulo; Ortoncelli, André Roberto; Pereira, Maria JoãoSkincare has become a constant demand among the population, who are increasingly concerned about their health. Furthermore, environmental issues arouse the interest of the masses in natural and sustainable products. This project proposes an approach for recommending thermal-based products based on a set of information provided by the user, combined with the results of computer vision algorithms (to identify the age and occurrence of wrinkles on the user’s forehead). A list of recommended products is generated Based on the profile determined for the user. To predict wrinkles, for each facial image sent by the user, we apply a pre-processing step that segments and prepares the region of interest, which a CNN will process. As a CNN, we used the VGG16 architecture trained using a transfer learning and fine-tuning strategy, which improved the results obtained, reaching an accuracy of 92% in classifying wrinkles. An algorithm provided by the Deepface tool is used to predict the user’s age, based on the sent picture.Which is crossed with the user’s information to determine a level of aging in order to improve the quality of the recommended products.
- Colorectal Polyp Segmentation: Impact of Combining Different Datasets on Deep Learning Model PerformancePublication . Araujo, Sandro Luis de; Scheeren, Michel Hanzen; Aguiar, Rubens Miguel Gomes; Mendes, Eduardo; Franco, Ricardo Augusto Pereira; Paula Filho, Pedro Luiz deColorectal cancer is a major health concern, ranking as one of the most common and deadly forms of cancer. It typically begins as polyps, which are abnormal growths in the intestinal mucosa. Identifying and removing these polyps through colonoscopy is a crucial preventative measure. However, even experienced professionals can overlook some polyps during examinations. In this context, segmentation algorithms can assist medical professionals by identifying areas in an image that correspond to a polyp. These algorithms, which rely on deep learning, require extensive image datasets to effectively learn how to identify and segment polyps. This study aimed to identify public colonoscopy image datasets that contain polyps and to examine how combining these datasets might affect the performance of a deep learning-based segmentation algorithm. After selecting the datasets and defining their combinations, we trained a segmentation algorithm on each combination. The evaluation of the trained models showed that merging datasets can enhance model generalization, with increases of up to 0.242 in the dice coefficient and 0.256 in the Intersection over Union (IoU). These improvements could lead to higher diagnostic accuracy in clinical settings, enhancing efforts to prevent colorectal cancer.
- Optimizing Olive Disease Classification Through Hybrid Machine Learning and Deep Learning TechniquesPublication . Mendes, João; Moso, Juliet; Berger, Guido; Lima, José; Costa, Lino; Guessoum, Zahia; Pereira, Ana I.Olive trees play a crucial role in the global agricultural landscape, serving as a primary source of olive oil production. However, olive trees are susceptible to several diseases, which can significantly impact yield and quality. This study addresses the challenge of improving the diagnosis of diseases in olive trees, specifically focusing on aculus olearius and Olive Peacock Spot diseases. Using a novel hybrid approach that combines deep learning and machine learning methodologies, the authors aimed to optimize disease classification accuracy by analyzing images of olive leaves. The presented methodology integrates Local Binary Patterns (LBP) and an adapted ResNet50 model for feature extraction, followed by classification through optimized machine learning models, including Stochastic Gradient Descent (SGD), Support Vector Machine (SVM), and Random Forest (RF). The results demonstrated that the hybrid model achieved a groundbreaking accuracy of 99.11%, outperforming existing models. This advancement underscores the potential of integrated technological approaches in agricultural disease management and sets a new benchmark for the early and accurate detection of foliar diseases.
- Predicting Retail Store Transaction Patterns: A Comparison of ARIMA and Machine Learning ModelsPublication . Vaz, Clara B.; Sena, Inês; Braga, Ana Cristina; Novais, Paulo; Lima, José; Pereira, Ana I.Retail transactions represent sales of consumer goods, or final goods, by consumer companies. This sector faces security challenges due to the hustle and bustle of sales, affecting employees’ workload. In this context, it is essential to estimate the number of customers who will appear in the store daily so that companies can dynamically adjust employee schedules, aligning workforce capacity with expected demand. This can be achieved by forecasting transactions using past observations and forecasting algorithms. This study aims to compare the ARIMA time series algorithm with several Machine Learning algorithms to predict the number of daily transactions in different store patterns, considering data variability. The study identifies four typical store patterns based on these criteria using daily transaction data between 2019 and 2023 from all retail stores of the leading company in Portugal. Due to data variability and the results obtained, the algorithm that presents the most minor errors in predicting daily transactions is selected for each store. This study’s ultimate goal is to fill the gap in forecasting daily customer transactions and present a suitable forecasting model to mitigate risks associated with transactions in retail stores.
- XAI Framework for Fall Detection in an AAL SystemPublication . Messaoudi, Chaima; Kalbermatter, Rebeca B.; Lima, José; Pereira, Ana I.; Guessoum, Zahia; Kalbermatter, Rebeca B.The Ambient Assisted Living (AAL) systems are humancentered and designed to prioritize the needs of elderly individuals, providing them with assistance in case of emergencies or unexpected situations. These systems involve caregivers or selected individuals who can be alerted and provide the necessary help when needed. To ensure effective assistance, it is crucial for caregivers to understand the reasons behind alarm triggers and the nature of the danger. This is where an explainability module comes into play. In this paper, we introduce an explainability module that offers visual explanations for the fall detection module. Our framework involves generating anchor boxes using the K-means algorithm to optimize object detection and using YOLOv8 for image inference. Additionally, we employ two well-known XAI (Explainable Artificial Intelligence) algorithms, LIME (Local Interpretable Model) and Grad-CAM (Gradient-weighted Class Activation Mapping), to provide visual explanations.