Computational Astronautical Sciences and Technologies (CAST)
The mission of the Oden Institute Computational Astronautical Sciences and Technologies (CAST) group is to assemble and lead the world’s top multidisciplinary science and technology research and development talent and focus it to solve problems in (a) astrodynamics, (b) space environment, (c) space propulsion, (d) satellite guidance, navigation, and control, and (e) spacecraft systems and design which require or substantially benefit from high performance computing and computational science capabilities. CAST’s intellectual core is at the intersection of astronautics, statistics and data science, computer science, and computational engineering.
CAST specifically seeks to integrate:
• Implementation of state-of-the-art theoretical approaches to multi-fidelity physics-based models and uncertainty quantification
• Development of advanced computational algorithms
• Use of large-scale computational frameworks
Part of the CAST mission is to deliver relevant astronautical models to the global community for public good. As such, CAST develops, leverages, and contributes to open-source software that can be found here: https://github.com/ut-astria
The vision of CAST is to imagine, identify, develop and deliver new computational astronautical sciences and technologies; make expertise on computational astronautical sciences and technologies available to a variety of stakeholders including all branches of government, private industry, academia, and international entities; and help guarantee the Oden Institute's leadership in the area through education, excellence, innovation and practicality in computational astronautical sciences and related fields of study.
The current areas of research of the CAST group are:
• Non-gravitational Astrodynamics
• Multi-body and Multi-Fidelity Uncertain Orbital Dynamical Systems Modeling
• Multi-source Information Fusion (including Hard [Physics-based] and Soft [Human-based] Inputs)
• Multi-sensor/target Tracking
• Machine/Deep Learning and clustering techniques
• Artificial Intelligence
• Semantic Reasoning and Natural Language Processing (NLP)
• Uncertainty Quantification, Assessment, and Prediction
• Low Signal-to-Noise Detection and Discovery
• Space Object and Event Taxonomies and Classification
• “Biometric” Methods for Unique Resident Space Object Identification and Characterization
• Spacecraft Guidance, Navigation, and Control (Autonomous and Ground Based)
• Space Traffic and Debris Modeling
• Space Object Aging (Gerontology) Effects and Impacts
• Break-up events, fatigue, cracking, fragmentation, sloughing
• Space Environment Effects and Impacts
• Near Earth Object Detection and Tracking
• Spacecraft In-Situ Data Exploitation
• Space Domain Awareness and Decision-Support Systems
Some examples of high-interest areas are:
Topic: Space Environment Effects and Impacts on Space Objects and Events
Application of physics-based algorithms and models leveraging high performance computing to quantify space object material aging/degradation, charging effects, and non-gravitational forces and torques for various classes of resident space objects (rocket bodies to multi-layer insulation). Realistic trajectory prediction of these objects is of interest.
Modeling and prediction of resident space object break-up events due to material fatigue, cracking, stress, etc.
Modeling and prediction of space object re-entries ad determining landing/survivability footprints and possible expected casualty calculations. This is highly computational and physics-based.
Topic: Resident Space Object Classification Leveraging Research Description Framework, Ontologies, and a “Biometric” Approach to Unique Identification
Investigate the utility of representing multi-source information in a RDF/Ontology-based Knowledge Graph and demonstrate the ability to use the RDF/Ontology as a method to classify (taxonomy) of man-made resident space objects. This would potentially have to be done for tens of thousands of space objects in the analyses. Also investigate the use of artificial intelligence and machine learning to apply to this classification problem.
Develop a method to uniquely identify resident space object based upon “biometric” methods.
Topic: Hard/Soft Information Fusion
Develop a method to combine and fuse both physics-based information (e.g. radar and telescopes) with human-based information(e.g. texts interrogated by Natural Language Processing). Investigate methods of Uncertainty Quantification and Representation for this problem and demonstrate realistic utility for several use-cases. How can one handle both probabilistic and opinion-based inputs in a common framework?
Topic: Use of Voronoi models for Uncertain Orbital Dynamical Systems
It is possible to represent uncertain orbital dynamical systems in the Voronoi framework?
In essence, orbital systems are modeled realistically as random variables and we can incorporate this uncertainty in the Voronoi framework
Is it possible for improved Space Object identification, classification, and behavioral prediction to be achieved by extending the parameter space in the Voronoi model to include physical and functional space object features, beyond solely kinematics states (i.e. position and velocity)?
Much like the molecular structure readily captured by the Voronoi theory, we proposed to define an equivalent “Space Object Molecule” whose parameter space includes orbital states but also space object characteristics such as mass, size, material properties, and perhaps even functional capabilities.
All of these efforts are supported by a multi-disciplinary blend of skills and expertise including foundational knowledge in applied mathematics, applied physics, numerical methods, probability and statistics, applied estimation theory, data models/engineering/science/analytics, and simulation algorithms. This is then harnessed and focused with complimentary domain-specific expertise in astrodynamics, spacecraft design and systems, sensor systems, space environment, spacecraft propulsion, and satellite guidance, navigation, and control.