Doutorado em Ciência da Computação
URI Permanente para esta coleção
Nível: Doutorado
Ano de início:
Conceito atual na CAPES:
Ato normativo:
Periodicidade de seleção:
Área(s) de concentração:
Url do curso:
Navegar
Submissões Recentes
- ItemGenerating road grid maps and path plans for self-driving cars using laser remission data and deep neural networks(Universidade Federal do Espírito Santo, 2024-09-24) Carneiro, Raphael Vivacqua; Souza, Alberto Ferreira de; https://orcid.org/0000-0003-1561-8447; Baduê, Claudine Santos; Rauber, Thomas Walter; Komati, Karin Satie; Andrade, Mariella BergerThis work proposes the use of deep neural networks (DNN) for solving the problem of inferring the location of drivable lanes of roadways and their relevant properties such as the lane change right-of-way, even if the line markings are poor or absent. This problem is relevant to the operation of self-driving cars which requires precise maps and precise path plans. Our approach to the problem is the use of a DNN for semantic segmentation of LiDAR remission grid maps into road grid maps. Both LiDAR remission grid maps and road grid maps are square matrices in which each cell represents features of a small 2D-squared region of the real world (e.g., 20cm × 20cm). A LiDAR remission grid map cell contains the information about the average intensity of laser reflection remission on the surface of that particular place. A road grid map cell contains the semantic information about whether it belongs to either a drivable lane or a line marking or a non-drivable area. The semantic codes associated with the road map cells contain all information required for building a network of valid paths, which are required for self-driving cars to build their path plans. Our proposal is a novel technique for the automatic building of viable path plans for self-driving cars. In our experiments we use the self-driving car of UFES, IARA (Intelligent Autonomous Robotic Automobile). We built datasets of manually marked road lanes and use them to train and validate the DNNs used for the semantic segmentation and the generation of road grid maps from laser remission grid maps. The results achieved an average segmentation accuracy of 94.7% in cases of interest. The path plans automatically generated from the inferred road grid maps were tested in the real world using IARA and has shown performance equivalent to that of manually generated path plans.
- ItemCrossing domains for accuracy: in-network stacking of machine learning classifiers(Universidade Federal do Espírito Santo, 2024-06-21) Xavier, Bruno Missi; Ruffini, Marco; https://orcid.org/0000-0001-6220-0065; Martinello, Magnos; https://orcid.org/0000-0002-8111-1719; Pacheco, André; Kirian, Mariam; Pasquini, Rafael; Aparicio, Albert CabellosTraffic management plays a crucial role for this expansive global connectivity. In this context, traffic classification strategically differentiates a range of applications and its requirements. This transformation enhances network agility and facilitates the direct integration of Machine Learning (ML) into the network infrastructure, fundamentally changing traffic management by promoting proactive data processing and analysis within the network. The synergy of Network Softwarization (NS) with ML not only leads to reduced latency and improved load management, but also increases the capacity to effectively handle larger volumes of data. Consequently, networks are better equipped to meet the complex demands of modern digital ecosystems, ensuring robust and efficient connectivity. This thesis delved into the domain of programmable networks, with a specific focus on implementing ML for traffic classification within the network architecture components, including the Radio Access Network (RAN) and programmable data planes. This approach represents a significant departure from the traditional traffic classification techniques, which are typically deployed on end-hosts. It also paves the way for integrating Cross Domain Artificial Intelligence (AI) capabilities within the network, facilitated by multi-view learning. More specifically, we advanced the state-of-the-art with four main contributions: (i) We design a framework named Early Attack Guarding and Learning Engine (EAGLE) as the first defense line against a set of cyber threats. Our framework explores Open Radio Access Network (O-RAN) to collect measurements from the air interface (that is, Physical and Medium Access Control) to identify and early mitigate malicious flows; (ii) We introduce MAP4 that demonstrates the feasibility of deploying ML models (that is, decision trees and Random Forest) in the data plane. To achieve this, we rely on the P4 language to deploy a pre-trained model to accurately classify flows at line rate; (iii) We proposed an In-Network Concept Drift to deal with the dynamic nature of the network traffic. This approach detects changes in traffic distribution by implementing Exponentially Weighted Moving Average (EWMA) overcoming the P4 limitations; (iv) Our Cross-Domain AI integrates multiple layers (RAN and programmable data planes) to form an in-network stacking of ML classifiers under the multi-view learning perspective. This innovative approach overcomes the challenges of a single layer in order to improve the overall accuracy of the classification system.
- ItemA planning pipeline for software systems of autonomous vehicles(Universidade Federal do Espírito Santo, 2023-09-04) Azevedo, Pedro Henrique Vieira de Oliveira; Gonçalves, Claudine Santos Baduê; https://orcid.org/0000-0003-1810-8581; http://lattes.cnpq.br/1359531672303446; http://lattes.cnpq.br/5958251435708396; Komati, Karin Satie; https://orcid.org/0000-0001-5677-4724; http://lattes.cnpq.br/9860697624155451; Santos, Thiago Oliveira dos; https://orcid.org/0000-0001-7607-635X; http://lattes.cnpq.br/5117339495064254; Boeres, Maria Claudia Silva; https://orcid.org/0000-0001-9801-2410; http://lattes.cnpq.br/0528154281423964; Wolf, Denis FernandoThe use of autonomous vehicles on public roads and in industrial environments has been growing in recent years and to achieve the higher level of autonomy, where no human intervention is needed, the vehicle must know the roads of the environment where it will operate, which allows it to travel from an initial location to a desired destination dealing with the uncertainties of the environments, such as static obstacles. In this work, we propose a planning pipeline for autonomous vehicle software systems. The pipeline is composed of an offline process and an online process. The offline process consists of two modules: (i) the Waypoints Editor, used to edit or even create entire new paths using an open-source vector graphics editor. (ii) The Multi-Level Road Network Generator constructs a representation of the environment in two levels of representation where the first one is used to guide the vehicle in the autonomous mission and the second one represents the paths and is used is route computation. The online process consists of three modules: (i) the Route Planner module to compute routes, (ii) the Off-Road Planner module to compute short paths in order to bring the car to the start of the route or to take the car from the end of the route to the final goal, and (iii) the Frenét Frames Path Planner to generate alternative paths to the right and left of the route and, in the presence of static obstacles, overtake them. We evaluated the performance of the proposed planning pipeline through simulations in three different real-world datasets using our Autonomous Vehicle Simulator module. Simulated experimental results showed that the proposed planning pipeline allowed the autonomous vehicle to know the environment as much as the user of the system wants and successfully execute missions from an initial position to a desired destination computing faster routes using the second level of the Multi-Level Road Network, dealing with static obstacles in the environment obeying the overtake safe distance restrictions imposed by the autonomous vehicle system.
- ItemUma plataforma para integração de dispositivos de saúde em sistemas de monitoramento remoto de pacientes(Universidade Federal do Espírito Santo, 2023-07-17) Celestrini, Jordano Ribeiro; Santos, Celso Alberto Saibel; https://orcid.org/0000000232875843; http://lattes.cnpq.br/7614206164174151; https://orcid.org/0000000327148593; http://lattes.cnpq.br/6705424425518474; Pimentel, Maria da Graça Campos; https://orcid.org/0000-0001-8264-5811; http://lattes.cnpq.br/4962820320879891; Gouvea, Sonia Alves; https://orcid.org/000000015180471X; http://lattes.cnpq.br/7268228122543743; Villaca, Rodolfo da Silva; https://orcid.org/0000000280513978; http://lattes.cnpq.br/3755692723547807; Aguilar, Paulo Armando CavalcanteThe increase in the number of people with a chronic health condition has led researchers from different areas to seek alternatives for monitoring these patients outside the hospital environment, in order to monitor them continuously and prevent the worsening of their health status. In this sense, one research topic rising is RPM (Remote Patient Monitoring). This approach allows monitoring the health condition of patients from a distance, helping to prevent new hospitalization episodes, and improving their quality of life, and the care administered by health providers. However, the currently available RPM solutions face some challenges, mainly in the integration of new devices. Usually, solutions are designed for a specific scenario and do not focus on problems related to the heterogeneity of devices available for health monitoring. A solution that provides means for the integration of new devices has the potential to collaborate in advancing the state of the art in research in this context. Therefore, this thesis proposes a solution to support the execution of RPM projects, composed of a reference architecture and its implementation, materialized as the HDash Remote Monitoring Platform. HDash implements the proposed conceptual architecture, providing means for new devices to be easily integrated in order to meet different application scenarios. To evaluate the approach, two case studies were conducted: the first monitoring chronic patients, followed for 20 months, and the second monitoring in remote locations, for 36 months. The studies allowed evaluation of several aspects of the proposed architecture and it is hoped, therefore, that this work can contribute to RPM and that new research in this context will benefit from the artifacts produced in this thesis.
- ItemDetecção automática de doenças em frutos do mamão a partir da análise de imagens por meio de redes neurais profundas(Universidade Federal do Espírito Santo, 2023-07-14) Moraes, Jairo Lucas de; Souza, Alberto Ferreira de; https://orcid.org/0000000315618447; http://lattes.cnpq.br/7573837292080522; https://orcid.org/0000-0002-5111-0811; http://lattes.cnpq.br/8743832227027911; França, Felipe Maia Galvão; https://orcid.org/0000-0002-8980-6208; http://lattes.cnpq.br/1097952760431187; Partelli, Fabio Luiz; https://orcid.org/0000000288300846; http://lattes.cnpq.br/6730543200776161; Komati, Karin Satie; https://orcid.org/0000-0001-5677-4724; http://lattes.cnpq.br/9860697624155451; Oliveira, Elias Silva de; http://lattes.cnpq.br/2210356035827181Horticulture plays an essential role in the economies of various countries, serving as a significant source of income and job creation, particularly in developing nations. Within this sector, papaya holds substantial importance, being cultivated in over 60 countries, including Brazil, which stands as the second-largest producer of this fruit. Papaya is a delicate and climacteric fruit, leading to considerable post-harvest losses, underscoring the pivotal role of early detection and accurate classification of fruit injuries in quality control and loss mitigation. Presently, papaya quality control is conducted manually, demanding exhaustive and repetitive efforts, often necessitating specialized knowledge that may not always be readily available to small-scale farmers or small fruit processing facilities. Given this backdrop, the implementation of autonomous or semi-autonomous system solutions aimed at assisting in papaya quality control, including disease detection and physical damage identification, is highly desirable. Such solutions could effectively mitigate industry losses, offering a more efficient, precise, and reliable approach to ensuring fruit quality and maximizing productivity in the sector. In this thesis, we propose a comprehensive solution spanning from the creation of an unprecedented dataset in the literature to the development of a mobile application. This includes the implementation of novel convolutional neural network (CNN) architectures utilizing the Convolutional Block Attention Module (CBAM). Our dataset comprises more than 23,000 examples of eight types of injuries (Anthracnose, Phytophthora, Chocolate Spot, Sticky Disease, Black Spot, Physiological Spot, Mechanical Damage, and Scar) affecting papaya fruits, alongside examples of healthy fruits. The developed detector achieves a new state-ofthe-art in papaya fruit disease detection, with an average precision (mAP) of 86.2%. This performance significantly surpasses that of human experts, who achieved an average precision of 67.3%. Lastly, we optimized the structure and weights of our detector for use on mobile devices and created a robust mobile application that can run on common smartphones. It can detect diseases in papaya fruits at a rate of up to 6 frames per second without requiring additional resources.