Rao-blackwellised particle filtering in random set multitarget tracking

Author

Vihola, Matti

Volume

43

Issue

2

fYear

2007

fDate

4/1/2007 12:00:00 AM

Firstpage

689

Lastpage

705

Abstract

This article introduces a Rao-Blackwellised particle filtering (RBPF) approach in the finite set statistics (FISST) multitarget tracking framework. The RBPF approach is proposed in such a case, where each sensor is assumed to produce a sequence of detection reports each containing either one single-target measurement, or a "no detection" report. The tests cover two different measurement models: a linear-Gaussian measurement model, and a nonlinear model linearised in the extended Kalman filter (EKF) scheme. In the tests, Rao-Blackwellisation resulted in a significant reduction of the errors of the FISST estimators when compared with a previously proposed direct particle implementation. In addition, the RBPF approach was shown to be applicable in nonlinear bearings-only multitarget tracking.

Keywords

Gaussian processes; Kalman filters; particle filtering (numerical methods); statistical analysis; target tracking; Rao-Blackwellised particle filtering; extended Kalman filter; finite set statistics; linear-Gaussian measurement model; multitarget tracking framework; nonlinear model; random set multitarget tracking; Approximation algorithms; Filtering; Monte Carlo methods; Particle tracking; Sampling methods; Sliding mode control; Solids; Statistics; Surveillance; Testing;

fLanguage

English

Journal_Title

Aerospace and Electronic Systems, IEEE Transactions on

Publisher

ieee

ISSN

0018-9251

Type

jour

DOI

10.1109/TAES.2007.4285362

Filename

4285362