Trabalho de Conclusão de Curso | Graduação

URI permanente desta comunidadehttps://repositorio.insper.edu.br/handle/11224/3244

Navegar

Resultados da Pesquisa

Agora exibindo 1 - 10 de 26

Identification of Flooding Incident Impacts Using Neural Networks
(2024) Santos, Alexandre Magno Maciel dos; Vaz, Eduardo Mendes; Cadorniga, João Lucas de Moraes Barros; Pertusi, Pedro Vaz de Moraes
In a world where flooding impacts are becoming increasingly common, such as the disaster in the Brazilian state of Rio Grande do Sul in 2024, the goal of this project is to develop an open-source flooding impact assessment pipeline. Preliminary technical evaluation by NVIDIA indicates that this tool could be integrated with technologies such as a flood simulation system, allowing for predictions in susceptible regions. The pipeline utilizes Convolutional Neural Networks (CNN), public population, and geographic data to process images extracted from the Sentinel-1 and Sentinel-2 satellites and generate metrics. This project classifies flooded regions prior to and after a crisis, providing, for example, estimations of the affected population by area to showcase the impact to assist urban planning professionals. The developed tool integrates a satellite data collection system from these satellites, as they are also open-source and include periodical data, and the CNNs in an intuitive and easily utilizable pipeline, inspired by UNOSAT’s Emergency Mapping which analyzed the impact of floodings in Nepal in 2021. Keywords:
Exploring RISC-V CPU for Aerospace Applications
(2024) Barreto, Arthur Martins de Souza; Barros, Eduardo Schneider Monteiro de; Patelli, Rodrigo Anciães; Assis, Victor Luis Gama de
This paper explores the potential of RISC-V CPUs for aerospace applications, focusing on the specific use case of Brushless Direct Current Motor (BLDC) control. The project aims to contribute to Brazil’s technological sovereignty by reducing reliance on foreign semiconductor technology. An initial implementation of a basic Six-Step algorithm demonstrated the feasibility of RISC-V for motor control. However, the limitations of this approach led to the investigation of more advanced Field- Oriented Control (FOC). While FOC offers superior performance, its implementation presented challenges related to timing constraints, current measurement accuracy, and rotor position feedback. The project successfully identified hardware requirements and constraints associated with BLDC motor control.
Estimating Room Temperature Arrival Time
(2024) Moura, Adney Costa; Drummond, Felipe Martins da Costa; Lopes, Lorran Caetano Machado; Alessi, Tomas Rodrigues
This project aims to develop a predictive model to estimate the temperature decay in a room equipped with air conditioning, using data collected by the Klima device, developed by Boldr, and integrating it into the company’s client environment for visualization. Klima is a device that integrates temperature and humidity sensors, allowing for both the control of air conditioning units and the transmission of data to the cloud, where it is stored and accessed through the company’s application. Based on this historical data, which includes variables such as the air conditioner’s operating mode and the temperature evolution over time, the project seeks to build a model capable of predicting the temperature decay curve of the environment, adapting to the specific behavior of each room. Besides developing the predictive model, the project also aims to integrate these predictions into Boldr’s application, making them accessible to customers, and to update the device’s firmware so that the predictions can also be displayed on a physical screen.
Dynamic Adaptation of Graphical User Interfaces in Augmented Reality Based on Environmental Factors: Enhancing User Experience and Safety
(2024) Santos, André Corrêa; Barão, Pedro Bittar; Lima, Rafael Melhado Araujo
Augmented Reality (AR) has been experiencing great steps in further blending real life and virtual environments, by bringing additional information and virtual controls to real world scenarios. As such, adequate and responsive positioning and integration of virtual elements are fundamental in AR to provide users a seamless experience in blending virtual elements to reality. This project aims to develop an Augmented Reality solution that makes graphical user interfaces (GUI), such as interactive panels, respond dynamically to their environment. The solution automatically adjusts the positioning and appearance of graphical interfaces based on conditions from the environment, such as changes in lighting, presence of important objects (such as warning signs), to the presence of people, and to the presence of possible safety risks to the user. To achieve this, multiple techniques in the area of computer vision have been used for identifying and classifying objects detected, while a Generative Artificial Intelligence (AI) model is used to interpret more nuanced contextual data, such as a user text input. This allows interfaces to adjust themselves to not block the vision of points of interest from the user, adjust colors and contrast for legibility and visual comfort. The development of the solution has been guided by user testing to ensure effectiveness and an intuitive experience. As of this report, a prototype has been developed that can adjust the positioning of the GUI to avoid occluding specific classes based on a text input. Additionally, it modifies the color of the GUI elements to complement the dominant colors behind GUIs in the camera image.
The Political Economy of Tyrannicide
(2024) Schwartsman, David Motta Nisencwajg
In this paper, we discuss tyrannicide as a rational political strategy and we evaluate its empirical effects on democratization. We first consider the murder of political leaders as a strategy inside basic selectorate competition, in selectorate theory, and we regard its equilibrium effects. We show that assassination risk can induce greater investment in public goods and pressure a reduction in coalition size and selectorate expansion. Then, using a staggered difference-indifferences strategy with Callaway and Santana’s reweighting estimator, we compare countries whose leaders survived murder attempts by chance with countries who leaders were killed. We observe evidence that the assassinations of dictators can favor democratization, increasing their countries polity-2 index by 3 to 4 points over time in the aggregate, with a lasting effect. We find weaker evidence for assassination of leaders in general.
Building NPCs for a real-time multiplayer game with Artificial Intelligence
(2024) Duarte, Diogo dos Reis; Cunha, Eduardo Araujo Rodrigues da; Barbosa, Letícia Coêlho; Domingos, Lídia Alves Chagas
This project developed NPCs (Non-Playable Characters) using Reinforcement Learning for the multiplayer game Arena of Dreams, developed by the company Fanatee, which blends the genres of Party Royale and Trivia, forming a game of various mini games in pursuit of the podium. The project's purpose is to construct an artificial intelligence model that enables an agent to behave like other players, capable of performing the required activities in the game, allowing the match to start even without the minimum required number of people. For the construction of the model, the Unity game engine was used in conjunction with the Unity Machine Learning Agents Toolkit (ML-Agents), which is an artificial intelligence agent system.
Development of an Administrator Web App for subscriptions’ management and a Mobile App for unlocking E-bikes
(2024) Ades, Cesar Ezra; Hadba, Lila Takahashi; Kawahara, Thiago Shiguero
This project aims to develop a Web Application (App) for the Administrator (E-Moving personnel) and an users’ Mobile App for E-Moving, a rental electric bike (E-Bike) company focused on improving urban mobility. This endeavour builds upon a prior Capstone Project (PFE), which developed a control board for electric bicycles in 2023. The current initiative seeks to enhance the E-Moving profitability and control, by being able to block users’ e-bikes and mitigating the risks of theft and default. The project, a collaboration between students from Insper (São Paulo, Brazil) and Texas A&M (Texas, United States), involves Insper students developing two Apps in accordance with the requirements established by the previous project and those of the Texas A&M students. Notably, the hardware component is being developed by Texas A&M students. The project is structured into five primary segments: (i) Web App and Mobile App screen flowchart, (ii) Web App and Mobile App screen design and front-end implementation, (iii) Bluetooth connection with E-bike, (iv) Database integration with both Web App and Mobile App front-end, and (v) Integration of Bluetooth, Web App, Mobile App, and E-bike. This project emphasizes a practical application that enables administrators (E-Moving personnel) to remotely monitor client's E-bikes and for them to manage its E-bike.
Development of a dashboard for real-time student performance data visualization
(2024) Carvalho, Arthur Ferreira; Carreras, Natália Queiroz Menezes; Mahfuz, Pedro Osborn
This project is designed to provide instructors who use the PrairieLearn platform with visual tools and insights into students’ performance by creating and displaying said information in an external dashboard. The data used to create the dashboard is pulled from PrairieLearn’s API. The objective with this is to help instructors model their courses to maximize academic engagement and performance. The PrairieLearn platform offers a dynamic and interactive environment for students to engage with course material, enabling instructors to create customizable quizzes, assignments, and assessments tailored to the evolving educational landscape. With their current system, which provides limited insights by a statistics table that displays average scores and completion times, the project aims to expand the analytical tools available to educators with the creation of this external tool. By integrating features such as performance metric analysis, question score histograms, and assessment completion percentage tracking, the dashboard equips instructors with a detailed view into student performance. This allows for a deeper understanding of assessment effectiveness, enabling educators in identifying learning gaps, adjusting teaching strategies, and customizing content to meet individual student needs more effectively. The anticipated outcome is a user-friendly dashboard, which provides insights into students' learning patterns, allowing instructors to make informed decisions and improve educational outcomes.
Search of shipwrecked people using drone swarms (part 2)
(2024) Oliveira, Jorás Custódio Campos de; Andrade, Pedro Henrique Britto Aragão; Falcão, Renato Laffranchi; Rodrigues, Ricardo Ribeiro
The project's purpose is to iterate on the given multi-agent Drone Swarm Search Environment (DSSE) and research into Reinforcement Learning methods. The DSSE was created with the direct purpose of using reinforcement learning algorithms to train swarms of drones to execute autonomous maritime search and rescue missions of shipwrecked people in the ocean. The environment simulates the movement of persons-in-water (PIW) considering the ocean's dynamic circumstances and calculates a dynamic map of probabilities to be given to the agents, with two distinct environments, one for rescue scenarios with simulated PIW and a second expanding on state-of-the-art research for maritime coverage search path planning. The DSSE facilitates the training and visualization of drone behavior, the project emphasizes continuous improvement and open accessibility, with the release of the DSSE as an open-source Python package and documentation. The focus is on the continuous improvement of simulation quality and applicability of the environments for research purposes, with development, training and evaluation of Reinforcement learning algorithms to improve the path planning of autonomous agents, for search and rescue maritime scenarios.
When labor-based parties gain businesses’ trust: the case of Brazil’s Workers’ Party
(2022) Lopes, Victor Hugo Almeida
Desde os processos de liberalização nas décadas de 80 e 90, elites financeiras e empresariais conseguem constranger partidos trabalhistas eleitos na América Latina a tentar buscar uma agenda econômica pró-mercado, assim comprometendo sua capacidade e incentivos de oferecer benefícios a seus eleitores tradicionais. No entanto, a implementação de políticas pró-negócios pelos partidos trabalhistas não depende apenas das pressões impostas pelos mercados internacionais de capitais, mas também é uma função da interação entre as elites partidárias e as suas bases. Assim, a forma como um partido se organiza desempenha um papel crucial nesse contexto, pois determina a alocação de barganha entre os dois. Meio a isso, este artigo, por meio de revisão bibliográfica e process-tracing, analisa o caso do Partido dos Trabalhadores no Brasil e as tentativas de Lula de conquistar a confiança do setor empresarial em seu primeiro mandato como presidente (2003-2006), comparando-o com outros casos semelhantes na América Latina. Como resultado, encontramos evidências de que, quando o partido do presidente tem fortes laços com movimentos sociais organizados, ao mesmo tempo em que concede autonomia às suas lideranças, o governo é mais propenso a optar pela ortodoxia de mercado, uma vez que as bases detêm menos poder.

Trabalho de Conclusão de Curso | Graduação

Navegar

Filtros

Configurações

Ordenar por

Resultados por página

Resultados da Pesquisa