Robust state estimation for small unmanned airplanes

As a basis for autonomous operation, Unmanned Aerial Systems (UAS) require an on-board state estimation that achieves both high accuracy as well as robustness with respect to certain conditions. We present a multi-sensor fusion framework based on Extended Kalman Filtering (EKF) which is light-weight enough to run on-board small unmanned airplanes using measurements from a MEMS based Inertial Measurement Unit (IMU), static and dynamic pressure sensors, as well as GPS (position and velocity) and a 3D magnetic compass. The on-board state estimator continuously estimates position, velocity, attitude and heading, IMU biases as well as the 3D wind vector in a tightly-coupled manner. In addition, airplane Angle of Attack (AoA) as well as sideslip angle can be derived by involving an aerodynamics model. The resulting infrastructure allows for unbiased orientation, airspeed and AoA tracking even in the case of GPS outages over extended periods of time. It can furthermore detect and reject outliers in sensor readings using Mahalanobis distance checks. We validate the proposed method with flight data from a small unmanned airplane: we demonstrate robustness w.r.t. outliers and GPS outages by disabling or corrupting respective flight data in post processing analyses.