Data Association Using Innovation Projections for Landmark-Based Localization

Mathieu Joerger, Ali Hassani, Matthew Spenko, and Jonathan Becker

Peer Reviewed

Abstract:	This paper presents a new data association method for bounding the integrity risk in landmark-based localization in ground transportation applications. Data association is the process of assigning currently-sensed landmark features to features that were previously observed or mapped. Most association methods use a nearest-neighbor criterion based on the normalized innovation squared (NIS). In contrast, we derive a new, closed-form, compact association criterion based on projections of the extended Kalman filter’s innovation vector. These innovation projections (IP) capture the impact of wrong associations on both the magnitude and direction of the innovation vector. We evaluate our newly derived IP method using simulated and experimental data for inertial-aided LiDAR localization in both indoor and outdoor environments. Compared to NIS, the proposed IP method (a) reduces the risk of wrong associations and (b) tightens the bound on predicted integrity risk.
Published in:	NAVIGATION: Journal of the Institute of Navigation, Volume 72, Number 4
Cite this article:	Citation Tools are available on the NAVIGATION open access site https://doi.org/10.33012/navi.715
Full Paper:	ION Members: Free Download Non-Members: Free Download Sign In