SEMINARIESCHEMA FÖR MATEMATISK STATISTIK 

Tisdag    11/6  14.00         Andrew Logothetis
          1996                Department of Electrical Engineering
                              University of Melbourne, Australia


                              Bearings-only target tracking using hidden
                              Markov models and an information theoretic
                              approach to observer path design for
                              bearings-only tracking
                              

Abstract:

Target motion analysis (TMA) is concerned with estimating parameters
associated with a target, namely it's position, velocity and other
information of interest. Some of these parameters are partially 
observed from noisy sets of measurements. When bearing information 
from a single moving platform to a target is used we have the problem 
of bearings-only tracking (BOT).

Generally in BOT a single target moves at constant velocity confined to
the same plane as the observer, which tows a linear array of hydrophones.
The observer passively monitors target bearings and it is desired to
obtain tight bounds on the location and velocity of this target, based
on the history of bearing measurements.

A new approach to BOT using hidden Markov models (HMMs) is presented.
Furtheremore, it is well known that the observer path influences system 
observability and the quality of the estimates. Results from information 
theory are used to design observer trajectories that maximize system 
observability and thus, minimize the uncertainty of the estimates.






Björn Holmquist 046-2228546