Sistema de monitorização e colisão entre tráfego espacial: projeto II
Date
2025
Journal Title
Journal ISSN
Volume Title
Publisher
IPCB. EST
Abstract
O tráfego espacial tem crescido rapidamente com o lançamento de satélites comerciais, científicos e militares. Este aumento tem vindo a aumentar o risco de colisões e a criação de detritos. Para manter operações sustentáveis, são necessárias soluções inteligentes que antecipem eventos de risco. Neste trabalho, foi realizada uma revisão da literatura sobre Deep Learning aplicado à monitorização orbital e à previsão de colisões. Observa-se o uso crescente de redes neuronais profundas, incluindo redes convolucionais (CNNs) e redes recorrentes (RNNs), para modelar trajetórias e encontros próximos. Neste contexto, iniciativas internacionais como o Spacecraft Collision Avoidance Challenge da Agência Espacial Europeia disponibilizam dados e métricas de referência que ajudam a comparar abordagens e a acelerar o desenvolvimento de soluções. O trabalho levou à obtenção, análise e preparação de datasets orbitais.
Foram avaliadas a estrutura, a qualidade e as limitações dos dados. O pré-processamento incluiu limpeza, normalização e seleção de atributos relevantes, deixando os dados prontos para uso em modelos de redes neuronais. Foi ainda desenvolvida uma aplicação web em Streamlit que permite explorar os dados, visualizar métricas e gráficos e experimentar protótipos de modelos de forma simples, servindo como prova de conceito e facilitando a análise interativa dos resultados.
Space traffic has grown rapidly with the launch of commercial, scientific and military satellites. This increase has raised the risk of collisions and the generation of debris. To maintain sustainable operations, intelligent solutions that anticipate risk events are required. In this work, we conducted a literature review on Deep Learning applied to orbital monitoring and collision prediction. There is a growing use of deep neural networks, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), model trajectories and close approaches. In this context, international initiatives such as the European Space Agency’s Spacecraft Collision Avoidance Challenge provide reference datasets and metrics that help compare approaches and accelerate the development of solutions. The project involved the acquisition, analysis and preparation of orbital datasets. We evaluated the structure, quality and limitations of the data. Preprocessing included cleaning, normalization and selection of relevant attributes, leaving the data ready for use with neural network models. We also developed a web application in Streamlit that enables exploration of the data, visualization of metrics and charts, and experimentation with model prototypes in a straightforward way, serving as proof of concept and facilitating interactive analysis of the results.
Space traffic has grown rapidly with the launch of commercial, scientific and military satellites. This increase has raised the risk of collisions and the generation of debris. To maintain sustainable operations, intelligent solutions that anticipate risk events are required. In this work, we conducted a literature review on Deep Learning applied to orbital monitoring and collision prediction. There is a growing use of deep neural networks, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), model trajectories and close approaches. In this context, international initiatives such as the European Space Agency’s Spacecraft Collision Avoidance Challenge provide reference datasets and metrics that help compare approaches and accelerate the development of solutions. The project involved the acquisition, analysis and preparation of orbital datasets. We evaluated the structure, quality and limitations of the data. Preprocessing included cleaning, normalization and selection of relevant attributes, leaving the data ready for use with neural network models. We also developed a web application in Streamlit that enables exploration of the data, visualization of metrics and charts, and experimentation with model prototypes in a straightforward way, serving as proof of concept and facilitating interactive analysis of the results.
Description
Keywords
Monitorização espacial, Prevenção de colisões, Inteligência artificial, Machine learning, Deep learning, Space monitoring, Collision prevention, Artificial intelligence
Citation
ESTEVES, Vitor Tomás Moreira ; RAMOS, André Alexandre Serra de Jesus (2025) - Sistema de monitorização e colisão entre tráfego espacial : projeto II. Castelo Branco: IPCB. EST. 93 p. Relatório do Trabalho de Fim de Curso de Engenharia Informática.