Izvorni znanstveni članak
https://doi.org/10.7305/automatika.2015.04.1051
Improved Pose Estimation for Vehicle Navigation using Frame Alignment and Forward Smoothing
Rok Juhant
orcid.org/0000-0003-3300-4983
; Dewesoft d.o.o Gabrsko 11a, 1240 Trbovlje, Slovenia
Darko Vrečko
; Department of Systems and Control, Jožef Stefan Institute, Jamova 39, SI-1000 Ljubljana, Slovenia
Jure Knez
; Dewesoft d.o.o Gabrsko 11a, 1240 Trbovlje, Slovenia
Sašo Blažič
orcid.org/0000-0002-9347-8534
; University of Ljubljana, Faculty of Electrical Engineering, Tržaška cesta 25, 1000 Ljubljana, Slovenia
Sažetak
In this paper we present the application of a modified Extended Kalman filter on a test device. This device is intended to be both rugged and simple to use, providing an accurate position and velocity output for land vehicles. As there are already many applications available for this purpose, our test device is unique in a way that it can be mounted in an arbitrary position on any metal surface on a vehicle, while it automatically discovers its orientation and aligns itself during the first stage of the test period. It comes as an extremely robust and low-cost solution. Furthermore, the pre-alignment outputs are corrected using a reverse output correction during the test period, immediately providing accurate outputs. The alignment algorithm greatly (by factor of 20 or more) reduces the initialization time. In addition, a novel smoothing algorithm with forward computation is described. The developed algorithm is tested with real-world experiments and proved to have a similar accuracy as the reference system, although using much cheaper and less-reliable sensor equipment.
Ključne riječi
Alignment; Extended Kalman Filter; GPS/INS Navigation; Smoothing
Hrčak ID:
152843
URI
Datum izdavanja:
20.10.2015.
Posjeta: 2.032 *