Monte Carlo Error Characterization of EKF-Based Image Aided Navigation

Daniel A. Marietta, Kenneth A. Fisher, Clark N. Taylor

Abstract:	Image Aided Navigation (IAN) is a self-contained navigation scheme that utilizes information contained in images taken of the scene surrounding the navigation vehicle to provide measurements that complement inertial data, reducing navigation errors. This paper presents the results of an effort to characterize the errors committed by a specific IAN algorithm. Such a characterization supports the further development of the algorithm by providing a baseline measure of expected filter performance against which future versions may be compared. To accomplish this task, a 100 run collection of inertial and imaging data was performed using a ground vehicle navigating a rectangular indoor path within a controlled environment. This data was processed offine through the prototype IAN algorithm proposed in [14], and a Monte Carlo error analysis performed of the position errors committed by the filter. Such an analysis of an IAN filter based upon a real world data collection of this size does not yet exist in the literature. Examples of filter divergence were seen within the analysis results, and these divergent runs were then examined separately, with an exploration of possible divergence indicators (filter computed covariance and measurement residuals) and suspected causes of the divergent behavior. The results of this work pro- vide a full characterization of the filter committed errors.
Published in:	Proceedings of the 2013 International Technical Meeting of The Institute of Navigation January 29 - 27, 2013 Catamaran Resort Hotel San Diego, California
Pages:	675 - 686
Cite this article:	Marietta, Daniel A., Fisher, Kenneth A., Taylor, Clark N., "Monte Carlo Error Characterization of EKF-Based Image Aided Navigation," Proceedings of the 2013 International Technical Meeting of The Institute of Navigation, San Diego, California, January 2013, pp. 675-686.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In