Abstract
Fixed wing Unmanned Aerial Vehicles (UAVs) are an increasingly common sensing platform, owing to their key advantages: speed, endurance and ability to explore remote areas. While these platforms are highly efficient, they cannot easily be equipped with air data sensors commonly found on their larger scale manned counterparts. Indeed, such sensors are bulky, expensive and severely reduce the payload capability of the UAVs. In consequence, UAV controllers (humans or autopilots) have little information on the actual mode of operation of the wing (normal, stalled, spin) which can cause catastrophic losses of control when flying in turbulent weather conditions. In this article, we propose a real-time air parameter estimation scheme that can run on commercial, low power autopilots in real-time. The computational method is based on a hybrid decomposition of the modes of operation of the UAV. A Bayesian approach is considered for estimation, in which the estimated airspeed, angle of attack and sideslip are described statistically. An implementation on a UAV is presented, and the performance and computational efficiency of this method are validated using hardware in the loop (HIL) simulation and experimental flight data and compared with classical Extended Kalman Filter estimation. Our benchmark tests shows that this method is faster than EKF by up to two orders of magnitude.