Conservative Uncertainty Estimation in Map-Based Vision-Aided Navigation

Zhen Zhu, Clark Taylor

Peer Reviewed

Abstract:	In a vision-aided autonomous system, it is crucial to have a consistent covariance matrix of the navigation solution. Overconfidence in covariance could lead to significant deviation of the navigation solution and failures of autonomous missions, especially in a GPS-denied environment. Consistency of a map-based, vision-aided navigation system is investigated in this paper. As has been shown in numerous previous works, the traditional extended Kalman filter (EKF) approach to navigation produces significantly inconsistent (overconfident) covariance estimates. Covariance Intersection (CI) and adjusted EKF approaches can both help resolve the over-confidence problem. We present both simulation-based and real-world results of each of these approaches and compare their strengths and weaknesses.
Published in:	Proceedings of the 2016 International Technical Meeting of The Institute of Navigation January 25 - 28, 2016 Hyatt Regency Monterey Monterey, California
Pages:	501 - 510
Cite this article:	Zhu, Zhen, Taylor, Clark, "Conservative Uncertainty Estimation in Map-Based Vision-Aided Navigation," Proceedings of the 2016 International Technical Meeting of The Institute of Navigation, Monterey, California, January 2016, pp. 501-510. https://doi.org/10.33012/2016.13451
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In