Interacting multiple model tracking using a neural extended Kalman filter

Author

Owen, Mark W. ; Stubberud, Stephen C.

Author_Institution

Orincon Corp., San Diego, CA, USA

Volume

4

fYear

1999

fDate

1999

Firstpage

2788

Abstract

We discuss the incorporation of the author´s neural extended Kalman filter (NEKF) into an interacting multiple model (IMI) architecture for use in multisensor target tracking. The NEKF is used in conjunction with a straight-line motion model. We compare this IMM implementation to a tracker with a straight-line motion model. The NEKF allows us to track through a manoeuvre with better precision than a straight-line motion model with high process noise because it is able to learn the manoeuvre online and improve the model prediction. Therefore, the NEKF better approximates the true dynamics of the target´s motion which improves the overall tracking performance

Keywords

Kalman filters; covariance matrices; filtering theory; neural nets; nonlinear filters; sensor fusion; state estimation; target tracking; interacting multiple model tracking; model prediction; multisensor target tracking; neural extended Kalman filter; overall tracking performance; straight-line motion model; Acceleration; Equations; Filters; Mathematical model; Motion measurement; Neural networks; Noise measurement; Predictive models; State estimation; Target tracking;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 1999. IJCNN '99. International Joint Conference on

Conference_Location

Washington, DC

ISSN

1098-7576

Print_ISBN

0-7803-5529-6

Type

conf

DOI

10.1109/IJCNN.1999.833522

Filename

833522