%0 Conference Proceedings
%F Brehard03a
%A Brehard, T.
%A Le Cadre, J.-P.
%A Prenat, M.
%A Cojan, Y.
%T Initialization of particle filter and posterior Cramer-Rao bound for bearings-only tracking in modified polar coordinate system
%B Symp. on Target Tracking and Sensor Data Fusion for Military Observation
%C Budapest, Hungary
%X We address the classical bearings-only tracking problem (BOT) for a singleobject which belongs to the class of non linear filtering problems.Recently, algorithms based on sequential Monte Carlo methods, also calledparticle filtering methods, have been proposed. However, wrong initialization in filtering algorithm has disastrous consequences on estimation quality. In this paper, modified polar coordinatesystem (MP) is proposed to solve this problem.This approach brings us to an interesting formulation of BOT problem withMP system. In particular, we show that this problem belongs to a more generalclass of problems: the non linear filtering problems with unknown statecovariance. We infer from this particular framework, a particle filteringalgorithm which can be initialized from a original method based on linearregression. Simulation results are presented for a typical tracking scenario, involving a manoeuvring observer and a constant velocity target. We study then how the posterior Cramer-Rao bound (PCRB) which provides a lower bound on the mean square error which can be obtained for BOT problem within the MP coordinate system. We propose original PCRB approximations for the ``partial'' state target (the observable components). Moreover, we show that PCRB for the ``full'' state target in MP coordinate system can be computed. In practice, it seems that these bounds are over optimistic. We argue that this is due to asymptotic bias assumptions and show how it is possible to remedy to this drawback
%8 October
%D 2003