Abstract
M.Ing. (Electrical And Electronic Engineering)
State estimation comprises the estimation of the position and velocity
(state) of a target based on the processing of noise-corrupted measurements
of its motion. This study views a class of measurement processes where the
states are unobservable and cannot be estimated without placing additional
constraints on the system. The bearings only target motion problem is
taken as being representative of this type of problem. The results of
this study indicate that practical state· estimation for systems with
unobservable measurement processes is possible with the application of
estimation theories and available estimation techniques. Due to the
inherent nonlinear geometrical characteristics the problem is classified as
a unobservable nonlinear estimation problem.
A review of state estimation and estimation techniques is presented. The
fundamental bearings only target motion concepts are discussed. A
representative selection of bearings only estimators made from the
published literature, is evaluated. The evaluation consists of a
theoretical analysis and a Monte Carlo simulation of the estimators. Two
realistic scenario's are considered. A classification framework is
presented which may be useful to practical engineers in selecting suitable
estimators.
Batch estimators are shown to be more stable and likely to be used in
bearings only applications than recursive estimators. The importance of
isolating the unobservable states from the observable states by using a
modified polar co-ordinate system, is stressed. It is also shown that
effective data processing can be achieved by using all available
measurements and a maximum likelihood estimator.