DocumentCode :
3221492
Title :
A constrained extended Kalman filter for target tracking
Author :
Nordsjo, Anders Erik
Author_Institution :
Saab Bofors Dynamics, Jarfalla, Sweden
fYear :
2004
fDate :
26-29 April 2004
Firstpage :
123
Lastpage :
127
Abstract :
An extended Kalman filter, EKF, is proposed for tracking the position and velocity of a moving target. The suggested method is based on a nonlinear model which, in addition, incorporates means for estimating possible nonlinearities in the measurements of the target position. In many practical scenarios, the initial estimates of target position and velocity deviate significantly from the true ones. In order to reduce the impact of erroneous initial conditions and, hence, obtain a faster initial convergence to an acceptable trajectory, a certain constrained form of the EKF, named the CEKF, is introduced. Although the original Kalman filter for a purely linear system is inherently stable, there is no guarantee that the linearized model used in the EKF gives a stable algorithm. Hence, it is interesting to note that the proposed CEKF under certain mild conditions renders an exponentially stable algorithm. It is shown that this latter method can conveniently be formulated as a nonlinear minimization problem with a quadratic inequality constraint.
Keywords :
Kalman filters; filtering theory; minimisation; parameter estimation; radar tracking; target tracking; tracking filters; constrained extended Kalman filter; exponentially stable algorithm; nonlinear minimization problem; position estimation; quadratic inequality constraint; target tracking; track-while-scan radar sensor; velocity estimation; Acceleration; Additive noise; Equations; Filters; Linear systems; Minimization methods; Noise measurement; Position measurement; Radar tracking; Target tracking;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Radar Conference, 2004. Proceedings of the IEEE
Print_ISBN :
0-7803-8234-X
Type :
conf
DOI :
10.1109/NRC.2004.1316407
Filename :
1316407
Link To Document :
بازگشت