Augmented Motion Models for Constrained Position Tracking with Kalman Filters

Beitrag bei einer Tagung

Details zur Publikation

Autor(en): Kautz T, Groh B, Eskofier B
Jahr der Veröffentlichung: 2016
Sprache: Englisch


Accurate position tracking is a crucial task in many applications ranging from car navigation over robot control to sports analysis. In order to improve the accuracy of position tracking, we introduce a novel method for constraining Kalman filters by incorporating prior knowledge in an augmented motion model. In contrast to previously reported methods, our approach does not require cumbersome tuning of additional filter parameters and causes less computational overhead. We demonstrate our method in the context of sports analysis in athletics. Using 34 data sets recorded during 400m and 800m runs, we compare our approach to unconstrained and pseudo-measurement filters.  The presented augmented motion model in conjunction with an Extended Kalman Filter (EKF) reduced the root mean square error of the filtered output by 60% compared to unconstrained filtering and by 50% compared to a pseudo-measurement EKF.

FAU-Autoren / FAU-Herausgeber

Eskofier, Björn Prof. Dr.
Stiftungs-Juniorprofessur für Sportinformatik (Digital Sports)
Groh, Benjamin
Stiftungs-Juniorprofessur für Sportinformatik (Digital Sports)
Kautz, Thomas
Stiftungs-Juniorprofessur für Sportinformatik (Digital Sports)


Kautz, T., Groh, B., & Eskofier, B. (2016). Augmented Motion Models for Constrained Position Tracking with Kalman Filters. Heidelberg, Germany.

Kautz, Thomas, Benjamin Groh, and Björn Eskofier. "Augmented Motion Models for Constrained Position Tracking with Kalman Filters." Proceedings of the 19th International Conference on Information Fusion (FUSION 2016), Heidelberg, Germany 2016.


Zuletzt aktualisiert 2018-10-08 um 06:54