| dc.description.abstract | An airborne Track?While?Scan (TWS) radar has an antenna mounted on top of the aircraft. The motion of the antenna limits the data rate of airborne TWS radars to very low values. Further, because of its large search volume, the number of targets that must be tracked by an airborne TWS radar is normally very high. Therefore, tracking in airborne radars is generally more complicated than tracking in ground?based radars. This calls for more efficient algorithms for airborne?radar data processing.
This thesis proposes numerically stable acceleration and jerk models in a target?oriented coordinate system for tracking highly manoeuvring targets using an extended Kalman filter. The proposed approach is well suited for airborne TWS radars, which typically have low data?update rates, long surveillance ranges, and added complexities because measurements are obtained from a moving platform. The algorithm described in this thesis is demonstrated to be superior, particularly with respect to bias errors and pseudo?acceleration at near ranges. During and after a manoeuvre, the filter is shown to settle faster towards the steady?state condition. Thus, it provides good smoothing under steady?state conditions and rapid manoeuvre following with minimal bias errors, even in low?data?rate systems.
Filtering and prediction methods are used to provide the best estimates of target kinematic quantities such as position, velocity, and acceleration. One of the major issues in the design of a tracking filter for airborne applications is the choice of the tracking coordinate system. While the measurements are made in the polar coordinate system, it is not directly used for tracking because of the pseudo?acceleration problem at near ranges. If a Cartesian coordinate system is used, measurements in polar coordinates need to be converted to Cartesian coordinates. In the process, measurements that are uncoupled in the polar system become coupled in the Cartesian system, requiring coupled covariance matrices for implementation.
In this thesis, an extensive study supported by computer simulation has been made to select a suitable tracking filter capable of tracking highly manoeuvring targets with minimum prediction and estimation errors.
A brief mathematical background of Kalman filters and the differences between tracking for airborne and ground?based radars are first presented. A study is then carried out to determine the choice of coordinate system for airborne TWS radars. It is shown that a target?oriented polar coordinate system is better suited for airborne?radar tracking.
The transition matrix for acceleration and jerk target models in the target?oriented coordinate system has been modified to include higher?order terms such as acceleration and the rate of change of acceleration, so that the filter is agile enough to track highly manoeuvring targets. The analysis of the filter was carried out using Monte Carlo runs by comparing estimation errors, prediction errors, and variance?reduction ratios. In all cases, the higher?order Kalman filter is found to perform better than the filters reported in the literature. | |
