Information-theoretic approach for concurrent path and sensor planning for a UAV with EO/IR sensors

This report presents an information-theoretic approach to concurrent path and sensor planning for a UAV equipped with gimballed EO/IR sensors. The goal is autonomuos UAV surveillance, i.e., search for objects along a road or in a certain area and localization of discovered targets. The work is inspired by research in optimal observer trajectory computations for bearings-only tracking. The path planning problem is first formulated and solved as an optimal control problem. A second method uses an interpolation method, e.g. splines, to solve the combined path and sensor planning problem. In the first method the optimization parameters are control signals for the UAV and in the second method the optimization parameters are waypoints of the UAV flight path and the sensor gazing directions. For both methods, an information matrix is constructed from the states representing targets and area points. A utility function is defined based on the information matrix and the planning process is to determine optimization parameters that maximize a utility function at the planning horizon. Different parameters and aspects of the methods are discussed, e.g. different utility functions, length of time horizon and prior information. The methods are applied to single target localization, n target localization, and area exploration. The spline method is also extended to handle a limited field-of-view sensor, as well as scene occlusion. Finally, simulation results of a road surveillance scenario are presented.