Autonomous search and surveillance with small fixed wing aircraft

Small unmanned aerial vehicles (UAVs) have the potential to act as low cost tools in a variety of both civilian and military applications including traffic monitoring, border patrol, and search and rescue. While most current operational UAV systems require human operators, advances in autonomy will allow these systems to reach their full potential as sensor platforms. This dissertation specifically focuses on developing advanced control, path planning, search, and image processing techniques that allow small fixed wing aircraft to autonomously collect data. The problems explored were motivated by experience with the development and experimental flight testing of a fleet of small autonomous fixed wing aircraft. These issues, which have not been fully addressed in past work done on ground vehicles or autonomous helicopters, include the influence of wind and turning rate constraints, the non-negligible velocity of ground targets relative to the aircraft velocity, and limitations on sensor size and processing power on small vehicles.; Several contributions for the autonomous operation of small fixed wing aircraft are presented. Several sliding surface controllers are designed which extend previous techniques to include variable sliding surface coefficients and the use of spatial vehicle dynamics. These advances eliminate potential singularities in the control laws to follow spatially defined paths and allow smooth transition between controllers. The optimal solution for the problem of path planning through an ordered set of points for an aircraft with a bounded turning rate in the presence of a constant wind is then discussed. Path planning strategies are also explored to guarantee that a searcher will travel within sensing distance of a mobile ground target. This work assumes only a maximum velocity of the target and is designed to succeed for any possible path of the target. Closed-loop approximations of both the path planning and search techniques, using the sliding surface controllers already discussed, are also studied. Finally, a novel method is presented to detect obstacles by segmenting an image into sky and non-sky regions. The feasibility of this method is demonstrated experimentally on an aircraft test bed.