Automated inference of impacting asteroids' physical properties and motion
Abstract/Contents
- Abstract
- In 2013, an unexpected asteroid of 20 m in diameter exploded in the air over Russia. This explosion led to over 1,000 injured civilians and millions of dollars spent to repair infrastructure damage. This small asteroid raised concerns about our current knowledge and characterization capabilities of asteroid populations. Our space assets and camera networks frequently observe impacts that can be leveraged to deepen our understanding of asteroid composition and origin if modelers simulate an event. However, manually modeling these events involves many trial-and-error variations to be evaluated, which is arduous because solutions are non-unique and dependent on case-specific modeling assumptions. The challenge of assessing a broad range of suitable matches results in the majority of detected impact events not being modeled. Typically, modeling occurs for events only when initial velocity, entry angle, and density measurements from meteorites are available because this information reduces the number of free parameters. These barriers to modeling require a systematic approach of evaluating and reporting matches. This proposed systematization reduces the current subjectivity in the quality of the fits and corresponding values of uncertain and poorly characterized parameters. There is a lot of data to analyze and, therefore, there is a need for tools to quickly and robustly characterize events. My research aims to solve this big data challenge by combining meteoroid science and engineering with artificial intelligence. This thesis introduces applications of artificial intelligence in asteroid risk assessment through the initial development of search-based optimization and supervised learning methods for inferring physical quantities based on observed energy deposition profiles. The search-based technique presented in this work was a genetic algorithm (GA) and the supervised learning approaches were random forest regressor, deep neural network, and convolutional neural network. In addition to the development of these proposed solutions, this development study addresses two central automation challenges: increasing the number of modeled impactors and generating realistic synthetic data. The automated modeling approaches provide human-level accuracy with increased alternate solutions and decreased human labor. The GA approach improves on previous modeling capabilities because human labor is tedious, consequently, leading to point estimates instead of reporting the possible range of solutions. On the other hand, the supervised learning approach provides point estimates with significantly less computational resources (time and memory) than the GA. Although the GA requires significantly higher run time requirements to provide solutions to one impact event than a pre-trained supervised learning method, the opportunity to produce a range of possible solutions provides more comprehensive results. The significance of this work is rooted in the initial development of two automated modeling techniques without extensive user-in-the-loop activity to enable the robust and quick characterization of asteroids' physical properties and motion in our atmosphere. This new capability enables meteor scientists to assess a large number of impactors at once, which is conducive to improved asteroid property distributions and uncertain modeling parameters for impact risk assessment.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2020; ©2020 |
| Publication date | 2020; 2020 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Tarano, Ana Maria |
|---|---|
| Degree supervisor | Close, Sigrid, 1971- |
| Thesis advisor | Close, Sigrid, 1971- |
| Thesis advisor | D'Amico, Simone |
| Thesis advisor | Kochenderfer, Mykel J, 1980- |
| Degree committee member | D'Amico, Simone |
| Degree committee member | Kochenderfer, Mykel J, 1980- |
| Associated with | Stanford University, Department of Aeronautics and Astronautics |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Ana María Tárano Gallardo. |
|---|---|
| Note | Submitted to the Department of Aeronautics and Astronautics. |
| Thesis | Thesis Ph.D. Stanford University 2020. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2020 by Ana Maria Tarano
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...