Publicação
Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit
| Resumo: | We have always been told since we were little that space is infinite. Having this in mind, it would not make much sense to be so cautious and aware of the space that lies above us. However, the area right above the Earth’s surface up to 2000 km is heavily contaminated with space debris which can have all kinds of origins and dimensions both man-made (inactive satellites, parts of rockets, minuscule flecks of paint) as well as from natural sources (small meteoroids). Considering that satellites have their propellant carefully measured to fulfill the planned trajectory and cannot afford evasion maneuvers at the slightest danger signal, it is important to quantify the uncertainty on the predictions made. To predict when two objects will collide, one will need to model their orbits with the goal of knowing their positions. Among the multiple elements involved, such as the gravity potential or the shape of the object, space weather is the most difficult to predict. Because of these stochastic variables, the early discoveries from multiple scientists in the eighteenth century were only enough to describe an orbit in the perfect case scenario. These variables make the modeling of a real orbit more challenging since they are random and have to be considered when modeling them since their effects are not negligible. One of the variables that has the most impact on calculating the orbit of a space object is atmospheric density. Since we are dealing with a physical system that abides by physical laws, even if not perfectly, this will be used to our advantage to improve the predictions made. As aforementioned, these laws known for centuries can be too tailored for the perfect-case scenario and new equations can be discovered to better model a real-case scenario. The objective of this research is to employ physically informed machine learning techniques for orbit determination as well as to model atmospheric density by leveraging physical domain knowledge and improving upon the standard approach. |
|---|---|
| Autores principais: | Funenga, João Pedro de Noronha |
| Assunto: | Physically-Informed Neural Networks Data-driven physical discovery Space Debris Orbital Mechanics Orbit determination |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| _version_ | 1868983637754511360 |
|---|---|
| author | Funenga, João Pedro de Noronha |
| author_facet | Funenga, João Pedro de Noronha |
| author_role | author |
| contributor_name_str_mv | Soares, Cláudia Guimarães, Marta Costa, Henrique RUN |
| country_str | PT |
| creators_json_txt | [{\"Person.name\":\"Funenga, João Pedro de Noronha\"}] |
| datacite.contributors.contributor.contributorName.fl_str_mv | Soares, Cláudia Guimarães, Marta Costa, Henrique RUN |
| datacite.creators.creator.creatorName.fl_str_mv | Funenga, João Pedro de Noronha |
| datacite.date.Accepted.fl_str_mv | 2023-12-01T00:00:00Z |
| datacite.date.available.fl_str_mv | 2024-03-05T18:41:04Z |
| datacite.date.embargoed.fl_str_mv | 2024-03-05T18:41:04Z |
| datacite.rights.fl_str_mv | http://purl.org/coar/access_right/c_abf2 |
| datacite.subjects.subject.fl_str_mv | Physically-Informed Neural Networks Data-driven physical discovery Space Debris Orbital Mechanics Orbit determination |
| datacite.titles.title.fl_str_mv | Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit |
| dc.contributor.none.fl_str_mv | Soares, Cláudia Guimarães, Marta Costa, Henrique RUN |
| dc.creator.none.fl_str_mv | Funenga, João Pedro de Noronha |
| dc.date.Accepted.fl_str_mv | 2023-12-01T00:00:00Z |
| dc.date.available.fl_str_mv | 2024-03-05T18:41:04Z |
| dc.date.embargoed.fl_str_mv | 2024-03-05T18:41:04Z |
| dc.format.none.fl_str_mv | application/pdf |
| dc.identifier.none.fl_str_mv | http://hdl.handle.net/10362/164463 |
| dc.language.none.fl_str_mv | eng |
| dc.rights.none.fl_str_mv | http://purl.org/coar/access_right/c_abf2 |
| dc.subject.none.fl_str_mv | Physically-Informed Neural Networks Data-driven physical discovery Space Debris Orbital Mechanics Orbit determination |
| dc.title.fl_str_mv | Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit |
| dc.type.none.fl_str_mv | http://purl.org/coar/resource_type/c_bdcc |
| description | We have always been told since we were little that space is infinite. Having this in mind, it would not make much sense to be so cautious and aware of the space that lies above us. However, the area right above the Earth’s surface up to 2000 km is heavily contaminated with space debris which can have all kinds of origins and dimensions both man-made (inactive satellites, parts of rockets, minuscule flecks of paint) as well as from natural sources (small meteoroids). Considering that satellites have their propellant carefully measured to fulfill the planned trajectory and cannot afford evasion maneuvers at the slightest danger signal, it is important to quantify the uncertainty on the predictions made. To predict when two objects will collide, one will need to model their orbits with the goal of knowing their positions. Among the multiple elements involved, such as the gravity potential or the shape of the object, space weather is the most difficult to predict. Because of these stochastic variables, the early discoveries from multiple scientists in the eighteenth century were only enough to describe an orbit in the perfect case scenario. These variables make the modeling of a real orbit more challenging since they are random and have to be considered when modeling them since their effects are not negligible. One of the variables that has the most impact on calculating the orbit of a space object is atmospheric density. Since we are dealing with a physical system that abides by physical laws, even if not perfectly, this will be used to our advantage to improve the predictions made. As aforementioned, these laws known for centuries can be too tailored for the perfect-case scenario and new equations can be discovered to better model a real-case scenario. The objective of this research is to employ physically informed machine learning techniques for orbit determination as well as to model atmospheric density by leveraging physical domain knowledge and improving upon the standard approach. |
| dirty | 0 |
| eu_rights_str_mv | openAccess |
| format | masterThesis |
| fulltext.url.fl_str_mv | https://run.unl.pt/bitstreams/0d78014b-597f-4f3a-a3b8-edae88f3f4a1/download |
| id | run_bcd9f8c4cebebedf4f0c8b7ee95be3c2 |
| identifier.url.fl_str_mv | http://hdl.handle.net/10362/164463 |
| inst_facet_str | urn:organizationAcronym:unl{{{_:::_}}}Universidade Nova de Lisboa |
| instacron_str | unl |
| institution | Universidade Nova de Lisboa |
| instname_str | Universidade Nova de Lisboa |
| language | eng |
| network_acronym_str | run |
| network_name_str | Repositório Institucional da UNL |
| oai_identifier_str | oai:run.unl.pt:10362/164463 |
| organization_str_mv | urn:organizationAcronym:unl |
| person_str_mv | Funenga, João Pedro de Noronha |
| publishDate | 2023 |
| repo_facet_str | urn:repositoryAcronym:run{{{_:::_}}}Repositório Institucional da UNL |
| reponame_str | Repositório Institucional da UNL |
| repository_id_str | urn:repositoryAcronym:run |
| service_str_mv | urn:repositoryAcronym:run |
| spelling | engpt_PTWe have always been told since we were little that space is infinite. Having this in mind, it would not make much sense to be so cautious and aware of the space that lies above us. However, the area right above the Earth’s surface up to 2000 km is heavily contaminated with space debris which can have all kinds of origins and dimensions both man-made (inactive satellites, parts of rockets, minuscule flecks of paint) as well as from natural sources (small meteoroids). Considering that satellites have their propellant carefully measured to fulfill the planned trajectory and cannot afford evasion maneuvers at the slightest danger signal, it is important to quantify the uncertainty on the predictions made. To predict when two objects will collide, one will need to model their orbits with the goal of knowing their positions. Among the multiple elements involved, such as the gravity potential or the shape of the object, space weather is the most difficult to predict. Because of these stochastic variables, the early discoveries from multiple scientists in the eighteenth century were only enough to describe an orbit in the perfect case scenario. These variables make the modeling of a real orbit more challenging since they are random and have to be considered when modeling them since their effects are not negligible. One of the variables that has the most impact on calculating the orbit of a space object is atmospheric density. Since we are dealing with a physical system that abides by physical laws, even if not perfectly, this will be used to our advantage to improve the predictions made. As aforementioned, these laws known for centuries can be too tailored for the perfect-case scenario and new equations can be discovered to better model a real-case scenario. The objective of this research is to employ physically informed machine learning techniques for orbit determination as well as to model atmospheric density by leveraging physical domain knowledge and improving upon the standard approach.application/pdfpt_PTPhysics-Inspired Machine Learning for orbit determination in Low-Earth OrbitFunenga, João Pedro de NoronhaSoares, CláudiaGuimarães, MartaCosta, HenriqueHostingInstitutionOrganizationalRUNe-mailmailto:run@unl.ptrun@unl.pt2024-03-05T18:41:04Z2023-122023-12-01T00:00:00ZHandlehttp://hdl.handle.net/10362/164463http://purl.org/coar/access_right/c_abf2open accessPhysically-Informed Neural NetworksData-driven physical discoverySpace DebrisOrbital MechanicsOrbit determination12778753 bytesliteraturehttp://purl.org/coar/resource_type/c_bdccmaster thesishttp://purl.org/coar/access_right/c_abf2application/pdffulltexthttps://run.unl.pt/bitstreams/0d78014b-597f-4f3a-a3b8-edae88f3f4a1/download |
| spellingShingle | Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit Funenga, João Pedro de Noronha Physically-Informed Neural Networks Data-driven physical discovery Space Debris Orbital Mechanics Orbit determination |
| status | SINGLETON |
| subject.fl_str_mv | Physically-Informed Neural Networks Data-driven physical discovery Space Debris Orbital Mechanics Orbit determination |
| title | Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit |
| title_full | Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit |
| title_fullStr | Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit |
| title_full_unstemmed | Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit |
| title_short | Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit |
| title_sort | Physics-Inspired Machine Learning for orbit determination in Low-Earth Orbit |
| topic | Physically-Informed Neural Networks Data-driven physical discovery Space Debris Orbital Mechanics Orbit determination |
| topic_facet | Physically-Informed Neural Networks Data-driven physical discovery Space Debris Orbital Mechanics Orbit determination |
| url | http://hdl.handle.net/10362/164463 |
| visible | 1 |