UBC Theses and Dissertations
Sensor management with applications in localization and tracking Ghassemi, Farhad
In this dissertation, we explore several themes in sensor management with an emphasis on their applications for target localization and tracking. We consider the sensor subset selection problem where a pre-specified number of sensors must be selected in a sensor network to estimate an unknown value of a time-invariant parameter, e.g. the position of a target. We study the Lagrangian and continuous relaxations of this problem with the determinant of the Fisher information matrix as the objective function. We prove that the continuous bound is tighter than the Lagrangian bound and outline an algorithm based on the so-called natural selection process to compute the continuous bound when sensors are allowed to make more than one measurement. We also study how a target can identify the informative sensors when it is facing a network that attempts to estimate its position or its other critical parameters. We show that by borrowing the notion of symmetric probabilistic values from cooperative game theory, the target can assign a power index to each sensor to determine how informative it is relative to the other ones. We further show that by choosing the determinant of the Fisher information matrix as the metric of estimation accuracy, the computational complexity associated with a power index gracefully increases with the number of sensors. Finally, we study the trajectory design problem for bearings-only tracking where the motion of a mobile sensor, called the observer, must be planned in order to estimate the position and the velocity of a moving target via bearing measurements. Our analysis of this problem demonstrates that the optimal solutions can be uniquely specified by only two ratios: (i) The distance that the observer can travel along a straight line during the observation period to the relative distance between the observer and the target. (ii) The speed of the observer relative to the speed of the target.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International