Uncertainty-aware control, planning, and learning for reliable robotic autonomy
Abstract/Contents
- Abstract
- As autonomous systems take on increasingly challenging tasks in safety-critical settings such as autonomous driving and aerospace, their ability to explicitly account for uncertainty becomes critical for ensuring their robustness. However, effectively accounting for uncertainty from various factors remains challenging. In particular, there is a pressing need for more efficient algorithms for (1) precisely propagating uncertainty through the decision making stack, (2) quickly selecting actions that optimize objectives and satisfy constraints under uncertainty, and (3) safely gathering informative data to adapt and reduce uncertainty. The first part of this thesis tackles the problem of reachability analysis, which entails characterizing all states that a dynamical system can reach. We introduce efficient sampling-based techniques and study their theoretical properties with random set theory, geometry, and optimal control, guiding the design of accurate reconstruction algorithms. We use these sampling-based methods to design robust control and planning algorithms. The second part of this thesis focuses on stochastic programming and trajectory optimization under uncertainty. We study the sample average approximation (SAA) approach and identify a limitation of the method for problems with expected value equality constraints. We formulate an alternative SAA, prove its asymptotic optimality under mild assumptions, and demonstrate it on challenging risk-averse trajectory optimization problems. The third part of this thesis presents a sequential exploration-exploitation framework for safe and active dynamics learning and control. This approach enables robots to autonomously learn their dynamics while satisfying constraints, until uncertainty levels are small enough to safely achieve desired tasks. The approach is validated on a robotic platform emulating spacecraft dynamics.
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 | 2023; ©2023 |
| Publication date | 2023; 2023 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Lew, Thomas Jonathan |
|---|---|
| Degree supervisor | Pavone, Marco, 1980- |
| Thesis advisor | Pavone, Marco, 1980- |
| Thesis advisor | Kochenderfer, Mykel J, 1980- |
| Thesis advisor | Schwager, Mac |
| Degree committee member | Kochenderfer, Mykel J, 1980- |
| Degree committee member | Schwager, Mac |
| Associated with | Stanford University, School of Engineering |
| Associated with | Stanford University, Department of Aeronautics and Astronautics |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Thomas Lew. |
|---|---|
| Note | Submitted to the Department of Aeronautics and Astronautics. |
| Thesis | Thesis Ph.D. Stanford University 2023. |
| Location | https://purl.stanford.edu/zk587cg3450 |
Access conditions
- Copyright
- © 2023 by Thomas Jonathan Lew
- 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...