Vision-Aided Nonlinear Observer for Fixed-Wing Unmanned Aerial Vehicle Navigation

Abstract

This paper presents a vision-aided uniformly semiglobally exponentially stable nonlinear observer for estimation of attitude, gyro bias, position, velocity, and specific force of a fixed-wing unmanned aerial vehicle, using measurements from an inertial measurement unit, a global navigation satellite system receiver, and computer vision. The computer vision uses optical flow from consecutive images from a camera together with the continuous epipolar constraint to calculate the scaled body-fixed linear velocity, namely, the direction of travel. Epipolar geometry eliminates dependency on the distance to objects in the images and the structure of the terrain being recorded, meaning there is no restriction on the types of terrain for which the observer is applicable. Experimental data from an unmanned aerial vehicle test flight and simulated data are presented, showing that the proposed nonlinear observer has robust performance. Experimental results are compared with an extended Kalman filter and illustrate that the estimates of the states converge to the correct values.