Real-Time Moving Horizon Estimation of Air Data Parameters and Wind Velocities for fixed-wing UAVs

We present a real-time implementation of an estimation algorithm for angle of attack, airspeed and wind velocities estimation on a single board computer. The estimator uses only sensor data from a standard fixed-wing UAV autopilot, which consists of a Global Navigation Satellite System receiver, an inertial measurement unit and a pitot-static tube. This sensor data is fused with a combination of kinematic, aerodynamic and stochstic wind models in a nonlinear moving horizon estimator using numerical optimization. An algorithmic differentiation toolbox and automatic code generation is used to create a real-time capable estimator which is able to run within a UAV on an on-board computer. Hardware in the Loop simulation results show that the latency of the estimator is significantly below the expected wind gust period and gives low root-mean-square estimation errors for angle of attack (0.29°), airspeed (0.21m/s) and wind velocities (0.44m/s).