{"id":{"repo_id":"mit","oai_identifier":"oai:dspace.mit.edu:1721.1/151812"},"canonical_url":"https://search.dev.ndltd.org/etd/mit/oai:dspace.mit.edu:1721.1/151812","repository":{"repo_id":"mit","name":"MIT","base_url":"https://dspace.mit.edu/oai/request"},"display":{"title":"Real-time Autonomy and Maneuvering Simulation of an Unmanned Underwater Vehicle Near a Moving Submarine Using Actively Sampled Gaussian Process Surrogate Models","abstract":"Unmanned Underwater Vehicle (UUV) maneuvering simulators have severe limitations on modeling UUV motion near a moving submarine because they are not capable of determining the complex, turbulent, hydrodynamic interactions in real time. Potential flow solvers are typically fast enough, but they neglect viscosity which introduces large inaccuracies that play a critical role in control. On the other hand, Computational Fluid Dynamics (CFD) accurately models these hydrodynamic interactions, but a simulation of a single UUV in one specific configuration typically takes hours or days to complete. Therefore, it is not practical for real-time applications. To bridge this gap, a machine learning framework based on actively sampled Gaussian Process (GP) regression is developed to create a reduced-order model (ROM) that predicts the hydrodynamic interactions in real time using a minimum number of expensive simulations. We show that the introduced active learning framework, called Non-Myopic MultiFidelity (NMMF) active learning for GP regression, significantly and parsimoniously accelerates the convergence of the surrogate model by combining the low cost of the low-fidelity, potential flow simulations to explore the domain, as well as optimally selected high-fidelity CFD simulations as training data to improve the model accuracy. It is shown that the resulting GP regression model captures accurately and efficiently the hydrodynamic interactions between the UUV and the moving submarine. Based on the developed algorithms, we are able to define operating envelopes that outline regions where the UUV safely overcomes the hydrodynamic interactions, as well as, regions where the UUV is overpowered and collides with the submarine. This approach also enables us to develop new autonomous protocols that compensate for the hydrodynamic interactions, by adjusting the desired UUV heading and speed, which enables the UUV to safely stay on the desired course. A sensitivity analysis confirms the robustness of the presented control strategies. The developed ideas pave the way for control algorithms in complex environments, such as turbulent boundary layers, which were previously impossible to navigate in real-time.","abstract_html":"Unmanned Underwater Vehicle (UUV) maneuvering simulators have severe limitations on modeling UUV motion near a moving submarine because they are not capable of determining the complex, turbulent, hydrodynamic interactions in real time. Potential flow solvers are typically fast enough, but they neglect viscosity which introduces large inaccuracies that play a critical role in control. On the other hand, Computational Fluid Dynamics (CFD) accurately models these hydrodynamic interactions, but a simulation of a single UUV in one specific configuration typically takes hours or days to complete. Therefore, it is not practical for real-time applications. To bridge this gap, a machine learning framework based on actively sampled Gaussian Process (GP) regression is developed to create a reduced-order model (ROM) that predicts the hydrodynamic interactions in real time using a minimum number of expensive simulations. We show that the introduced active learning framework, called Non-Myopic MultiFidelity (NMMF) active learning for GP regression, significantly and parsimoniously accelerates the convergence of the surrogate model by combining the low cost of the low-fidelity, potential flow simulations to explore the domain, as well as optimally selected high-fidelity CFD simulations as training data to improve the model accuracy. It is shown that the resulting GP regression model captures accurately and efficiently the hydrodynamic interactions between the UUV and the moving submarine. Based on the developed algorithms, we are able to define operating envelopes that outline regions where the UUV safely overcomes the hydrodynamic interactions, as well as, regions where the UUV is overpowered and collides with the submarine. This approach also enables us to develop new autonomous protocols that compensate for the hydrodynamic interactions, by adjusting the desired UUV heading and speed, which enables the UUV to safely stay on the desired course. A sensitivity analysis confirms the robustness of the presented control strategies. The developed ideas pave the way for control algorithms in complex environments, such as turbulent boundary layers, which were previously impossible to navigate in real-time.","abstract_has_math":false,"creators":["Hammond, Brady M."],"institution":"Massachusetts Institute of Technology","degree_name":"Doctoral","degree_level":null,"degree_discipline":null,"degree_department":"Massachusetts Institute of Technology. Department of Mechanical Engineering","school":null,"contributors":[],"advisors":["Sapsis, Themistoklis P."],"committee_chairs":[],"committee_members":[],"year":2023,"date_issued":"2023-06","date_published":"2023-06","updated_at":"2026-07-22T22:21:07Z","subjects":[],"languages":[],"rights":["In Copyright - Educational Use Permitted","Copyright retained by author(s)"],"rights_urls":["https://rightsstatements.org/page/InC-EDU/1.0/"],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/1721.1/151812","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Sapsis, Themistoklis P."]},{"key":"dc:contributor.department","label":"Department","values":["Massachusetts Institute of Technology. Department of Mechanical Engineering"]},{"key":"dc:creator","label":"Author","values":["Hammond, Brady M."]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2023-08-23T16:10:37Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2023-08-23T16:10:37Z"]},{"key":"dc:date.issued","label":"Date","values":["2023-06"]},{"key":"dc:publisher","label":"Institution","values":["Massachusetts Institute of Technology"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Doctoral","Doctor of Philosophy"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["In Copyright - Educational Use Permitted","Copyright retained by author(s)"]},{"key":"dc:rights.uri","label":"Rights URI","values":["https://rightsstatements.org/page/InC-EDU/1.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1721.1/151812"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Unmanned Underwater Vehicle (UUV) maneuvering simulators have severe limitations on modeling UUV motion near a moving submarine because they are not capable of determining the complex, turbulent, hydrodynamic interactions in real time. Potential flow solvers are typically fast enough, but they neglect viscosity which introduces large inaccuracies that play a critical role in control. On the other hand, Computational Fluid Dynamics (CFD) accurately models these hydrodynamic interactions, but a simulation of a single UUV in one specific configuration typically takes hours or days to complete. Therefore, it is not practical for real-time applications. To bridge this gap, a machine learning framework based on actively sampled Gaussian Process (GP) regression is developed to create a reduced-order model (ROM) that predicts the hydrodynamic interactions in real time using a minimum number of expensive simulations. We show that the introduced active learning framework, called Non-Myopic MultiFidelity (NMMF) active learning for GP regression, significantly and parsimoniously accelerates the convergence of the surrogate model by combining the low cost of the low-fidelity, potential flow simulations to explore the domain, as well as optimally selected high-fidelity CFD simulations as training data to improve the model accuracy. It is shown that the resulting GP regression model captures accurately and efficiently the hydrodynamic interactions between the UUV and the moving submarine. Based on the developed algorithms, we are able to define operating envelopes that outline regions where the UUV safely overcomes the hydrodynamic interactions, as well as, regions where the UUV is overpowered and collides with the submarine. This approach also enables us to develop new autonomous protocols that compensate for the hydrodynamic interactions, by adjusting the desired UUV heading and speed, which enables the UUV to safely stay on the desired course. A sensitivity analysis confirms the robustness of the presented control strategies. The developed ideas pave the way for control algorithms in complex environments, such as turbulent boundary layers, which were previously impossible to navigate in real-time."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["Ph.D."]},{"key":"dc:title","label":"Title","values":["Real-time Autonomy and Maneuvering Simulation of an Unmanned Underwater Vehicle Near a Moving Submarine Using Actively Sampled Gaussian Process Surrogate Models"]}]}],"canonical_facts":{"dc:contributor.advisor":["Sapsis, Themistoklis P."],"dc:contributor.department":["Massachusetts Institute of Technology. Department of Mechanical Engineering"],"dc:creator":["Hammond, Brady M."],"dc:date.accessioned":["2023-08-23T16:10:37Z"],"dc:date.available":["2023-08-23T16:10:37Z"],"dc:date.issued":["2023-06"],"dc:description.abstract":["Unmanned Underwater Vehicle (UUV) maneuvering simulators have severe limitations on modeling UUV motion near a moving submarine because they are not capable of determining the complex, turbulent, hydrodynamic interactions in real time. Potential flow solvers are typically fast enough, but they neglect viscosity which introduces large inaccuracies that play a critical role in control. On the other hand, Computational Fluid Dynamics (CFD) accurately models these hydrodynamic interactions, but a simulation of a single UUV in one specific configuration typically takes hours or days to complete. Therefore, it is not practical for real-time applications. To bridge this gap, a machine learning framework based on actively sampled Gaussian Process (GP) regression is developed to create a reduced-order model (ROM) that predicts the hydrodynamic interactions in real time using a minimum number of expensive simulations. We show that the introduced active learning framework, called Non-Myopic MultiFidelity (NMMF) active learning for GP regression, significantly and parsimoniously accelerates the convergence of the surrogate model by combining the low cost of the low-fidelity, potential flow simulations to explore the domain, as well as optimally selected high-fidelity CFD simulations as training data to improve the model accuracy. It is shown that the resulting GP regression model captures accurately and efficiently the hydrodynamic interactions between the UUV and the moving submarine. Based on the developed algorithms, we are able to define operating envelopes that outline regions where the UUV safely overcomes the hydrodynamic interactions, as well as, regions where the UUV is overpowered and collides with the submarine. This approach also enables us to develop new autonomous protocols that compensate for the hydrodynamic interactions, by adjusting the desired UUV heading and speed, which enables the UUV to safely stay on the desired course. A sensitivity analysis confirms the robustness of the presented control strategies. The developed ideas pave the way for control algorithms in complex environments, such as turbulent boundary layers, which were previously impossible to navigate in real-time."],"dc:description.degree":["Ph.D."],"dc:identifier.uri":["https://hdl.handle.net/1721.1/151812"],"dc:publisher":["Massachusetts Institute of Technology"],"dc:rights":["In Copyright - Educational Use Permitted","Copyright retained by author(s)"],"dc:rights.uri":["https://rightsstatements.org/page/InC-EDU/1.0/"],"dc:title":["Real-time Autonomy and Maneuvering Simulation of an Unmanned Underwater Vehicle Near a Moving Submarine Using Actively Sampled Gaussian Process Surrogate Models"],"dc:type":["Thesis"],"thesis:degree_name":["Doctoral","Doctor of Philosophy"]},"updated_at":"2026-07-22T22:21:07Z"}