This paper presents a simple data-driven approach to improve ground target tracking by an unmanned aerial vehicle (UAV) for certain classes of target trajectories from learned local linear models. The UAV is assumed to be a small fixed-wing aircraft equipped with a gimbaled camera for visual sensing. We attempt to build a controller from measurement data by building an augmented Linear Quadratic Regulator (LQR) system from an approximated linear operator that indirectly captures the properties of the target system. We evaluate the relative performance improvement gained by this data-driven approach versus the standard target following LQR system and provide bounds for this improved performance. We also consider the effect of sensor noise on the tracking performance resulting from the noisy and erroneous datasets. We demonstrate the performance of this method on a range of numerical data representing different classes of target trajectories.

This content is only available via PDF.
You do not currently have access to this content.