High-integrity urban localization : bringing safety in aviation to autonomous driving
Abstract/Contents
- Abstract
- Localization is a crucial function for autonomous vehicles (AVs), as it influences critical tasks such as perception, planning, and control. However, most existing localization systems focus on accuracy and may fail unexpectedly in challenging conditions, posing safety risks. To address this issue, this thesis proposes novel algorithms that prioritize safety along with accuracy for localization. These algorithms not only robustly estimate the location, but also continuously assess its reliability. Drawing inspiration from the principles of integrity monitoring -- a safety-centric framework used in aviation for reliable Global Navigation Satellite System (GNSS)-based localization -- this thesis aims to adapt these principles for assessing location reliability in autonomous driving. However, directly applying this framework to autonomous driving, particularly in urban environments, presents significant challenges. Unlike the well-characterized GNSS errors observed in open-sky conditions of aviation, localization errors in urban settings are more unpredictable and severe. Moreover, the variety of sensors employed in AV localization, such as GNSS, cameras, and inertial sensors, require effectively handling the limitations and combined effects of each sensor on the localization process. In this thesis, we present novel algorithms that utilize different sensors to improve localization and enable integrity monitoring in urban environments. These algorithms are tailored to handle various challenges that affect the reliability of the localization process, such as large errors in multiple GNSS measurements, high-dimensional errors in visual measurements, and combined effects of errors from different sensors. For these challenges, we develop strategies to mitigate the impact of large errors on localization, estimate the error bounds of the localization solution, or detect when the localization error exceeds a threshold. We also describe how these algorithms can be implemented efficiently for fast computation in AVs and show how they provide a better understanding of the localization reliability than existing methods. We test our algorithms on simulated and real-world data from urban environments. The results show that our algorithms can estimate the vehicle's position reliably and robustly, and also evaluate the integrity of the localization solution.
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 | Gupta, Shubh |
|---|---|
| Degree supervisor | Gao, Grace X. (Grace Xingxin) |
| Thesis advisor | Gao, Grace X. (Grace Xingxin) |
| Thesis advisor | Lall, Sanjay |
| Thesis advisor | Weissman, Tsachy |
| Degree committee member | Lall, Sanjay |
| Degree committee member | Weissman, Tsachy |
| Associated with | Stanford University, School of Engineering |
| Associated with | Stanford University, Department of Electrical Engineering |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Shubh Gupta. |
|---|---|
| Note | Submitted to the Department of Electrical Engineering. |
| Thesis | Thesis Ph.D. Stanford University 2023. |
| Location | https://purl.stanford.edu/hs838yq9876 |
Access conditions
- Copyright
- © 2023 by Shubh Gupta
- 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...