UAV Sensor FDI in Duplex Attitude Estimation Architectures Using a Set-Based Approach

A fault detection and isolation algorithm for the attitude estimation of an unmanned aerial vehicle (UAV) using low-cost magnetometers, accelerometers, and gyroscopes, implemented in an inertial measurement unit (IMU) is proposed. Assuming the availability of double triaxial gyros, accelerometers, and magnetometers, the possibility that fault can occur both in input (gyros outputs) and in output (accelerometers and magnetometers outputs) to the kinematic nonlinear relations underlying attitude estimation must be taken into account. If an extended Kalman filter (EKF) can compensate for biases on gyroscopes, fault detection, for all sensors, is first obtained with a comparison between homogeneous sensor outputs. Then, the isolation of the faulted gyro is carried out through the analysis of the EKF bias estimates, whereas a set-based approach is used to isolate faulted accelerometers or magnetometers. For these sensors, isolation tests involve the solution of a linear programming problem on a moving time window in the discrete time domain. In order to show the practical applicability and robustness against measurement noise and different kind of faults, a set of simulations involving experimental data collected during flights of a tricopter UAV are discussed.