Cimmino, Nicola (2024) Event, Object And Mission Characterization for Space Domain Awareness. [Tesi di dottorato]

Anteprima

Testo Cimmino_Nicola_36.pdf Download (6MB) | Anteprima

Abstract

This PhD dissertation discusses the design, development and testing of innovative algorithms, methods and tools applicable in the context of Space Situational Awareness. Three main research topics have been addressed: fragmentation events characterization; use of Machine Learning approaches for ballistic coefficient estimation based on astrometric data; optimal manoeuvring strategy for responsive space applications. In the context of Space Surveillance and Tracking (SST), operations such as detection, monitoring, and characterization of space objects are crucial for various functions, such as accurate orbit determination and collision risk assessment. In recent years, growing interest has been shown for space debris due to the lack of knowledge about their physical characteristics and thus the inaccurate determination of their orbits. To address this, a first research area of this PhD activity focused on improving the characterization of in-orbit fragmentation events, which represent a significant source of space debris. Specifically, starting from the implementation of the NASA's Standard Breakup Model, a primary contribution of the research activity was to develop an iterative approach that improves the estimation of the fragmented mass of two space objects colliding in orbit, leveraging the number of catalogued fragments after a certain time from the fragmentation event. The second research path involved the development of Machine Learning (ML) approaches for estimating the ballistic coefficient (BC) of resident space objects in Low Earth Orbit, exploiting astrometric data. Various ML techniques were explored, particularly two types of neural network architectures: Feedforward Neural Networks and Recurrent Neural Networks. In both cases, the sensitivity of these architectures to the quantity and quality of data was analysed, especially robustness to measurement errors. Additionally, the effects of solar activity and atmospheric model uncertainty on BC estimates were evaluated. The ML approach was then compared with conventional characterization techniques. Finally, in the complex and congested space environment, the ability of space assets to react to unforeseen situations is essential. In this context, an optimal multi-satellite manoeuvring strategy was proposed, to ensure the observation of areas of interest on the Earth's surface. Moreover, tools to support several Recognised Space Picture functions were developed, such as visibility analysis from ground-based sensors as well as GNSS coverage.

Downloads

Actions (login required)

Modifica documento