Abstract
M.Ing.
The results of this study indicate that developments in
tracking filter theory and microprocessor technology
during the past two decades have been responsible for more
accurate real-time state estimation and prediction. An
overview of tracking filters and p~predictors is presented.
A rep~esentative selection of tracKing filters and
p~edictors drawn from literature publ ished over the past
two decades, was reviewed. The selection consisted of
adaptive,
constant
non-adaptive, second and third order filters. A
velocity, constant acceleration and correlated
acceleration predictors are evaluated.
The tracking filter and position prediction evaluation
consisted of a theoretical analysis of tracking filters
and a Monte-Carlo simulation of tracking filters and
predictors for a one dimensional trajectory and two
real istic target trajectories.
The outstanding
tracKing fil ters
characteristics of the
are their relatively
second order
ideal dynamic
behaviour and an exceptional sensitivity to target
accelerations. Excessive transients and the absence of lag
for constant acceleration ta~gets characterise third order
filters. The choice of an optimal filter-predictor
combination is shown to be strictly appl ication dependent