Scalable System Approaches to Unmanned Aerial Vehicle Upset Prevention and Recovery

Wind shear, icing, wake vortices from ship superstructures or other aircraft, actuator malfunctions, or component failures can all contribute to upset conditions. For piloted aircraft, prevention of or recovery from these events is challenging because of the nonlinear dynamics encountered at angles of attack and sideslip outside of the normal flight envelope. The problems are magnified for unmanned aircraft given typical vehicle sizes, actuator bandwidth, and the absence of a pilot. The goal of our research is to develop the control technologies that enable unmanned aerial vehicles to perform flight-envelope protection and upset recovery autonomously. Reinforcement learning control is a core technology to design outer-loop controllers that affect situation-appropriate recovery within the problem constraints (structural loads, for example). Manual recovery practices and NATOPS procedures are also encoded in the design. Novel control devices can make air vehicles more resistant to departure by postponing the onset of flow separation at high angles of attack. Thus, we address the role of add-on actuators in the recovery control-system framework. These range from a flip-tail for a flight-test UAV to arrays of synthetic jets for a tailless UCAV. High-fidelity simulations of two dynamically dissimilar models are used to develop the technology in Phase I.