|
Abstract:
|
This thesis was made in cooperation with the Finnish Air Force and the Digital and Computer Systems Laboratory at Tampere University of Technology. Tracking is the process of estimating the current state of a target of interest using measurements obtained from it. In this thesis, attention was paid on flying targets only. Two filtering algorithms, namely the unscented Kalman filter and the iterated extended Kalman filter were implemented and their performance was compared to the first and second order Kalman filters. A simulation scenario was created and the tracking was performed using three different kinematic models. The performance of the algorithms on different trajectories was compared in the terms of tracking accuracy and computation time. Information was available only from two passive, non-moving sensors. The unscented Kalman filter had previously been tested in problems that contained significant nonlinearities. In this work, the filter was studied in a state estimation problem, where the state prediction is linear. The filter was found efficient especially when tracking highly manoeuvring or stationary targets. On other kinds of targets, the second order extended Kalman filter proved to be a slightly better choice. However, the unscented Kalman filter beat the first order extended Kalman filter in tracking accuracy. The iterated extended Kalman filter succeeded to track only the stationary target. /Kir10 |