Dead Reckoning of a Fixed-Wing UAV with Inertial Navigation Aided by Optical Flow

Abstract

This paper provides experimental results for dead reckoning of a fixed-wing UAV using a nonlinear observer (NLO) and a more recent tool called eXogenous Kalman Filter (XKF), which uses the NLO itself as a first-stage filter. The sensors used are an IMU (accelerometers, inclinometers, and rate gyros), a camera, and an altimeter; the observed states are position, velocity, and attitude. A machine vision system provides the body-fixed velocity of the UAV. Although the calculated velocity results affected by a bias, it is necessary both for estimating the attitude and for bounding the rate of divergence of the position during dead reckoning. Gyro, accelerometer, and optical flow (OF) velocity biases are estimated, but only as long as GNSS is available. When dead reckoning begins, they are frozen at their last calculated value. The experimental results show that the position error grows at a bounded rate with the proposed estimators.