Loss of control (LOC) due to upset is one of the main causes of accidents in manned aircraft and is already emerging as a significant causal factor in unmanned aircraft accidents. On manned aircraft, recovery from an upset condition relies on the skill and training of an expert pilot. Due to reduced situational awareness and delays introduced by the command and control link, it is unlikely that a remote UAS operator will be able to serve this function. An advanced system capable of perception, cognition, and decision making is necessary to replace the need for an operator with upset recovery expertise and to mitigate the LOC risk on UAS. Barron Associates has recently developed a two-stage architecture that generates safe and effective recovery maneuvers for a large set of upset conditions including full stall and fully-developed spin modes. The proposed research will design an upset detection system and integrate the system with the existing two-stage recovery architecture to yield a comprehensive autonomous upset recovery system. The decision about when to activate each stage of a recovery is difficult to make at design-time due to insufficient aerodynamic data and the inability to model all of the off-nominal precipitating factors that cause upsets. The proposed upset detection system does not rely on design-time characterization; instead, a rigorous statistical testing framework combines numerous pieces of information including vehicle attitude, rotational rate, and controller performance to answer the question: Has an upset occurred? Key Phase I goals include: upset detection algorithm development, integration of upset detection with existing recovery architecture, evaluation of system performance in simulation, and real-time hardware-in-the-loop demonstration using a commercially available autopilot.