Relative navigation of fixed‐wing aircraft in GPS‐denied environments

This work enables GPS‐denied flight on fixed‐wing UAS by accounting for fixed‐wing‐specific sensing requirements and using a methodology called relative navigation as an overarching framework. The development of an odometry‐like, front‐end, EKF‐based estimator that utilizes only a monocular camera and an inertial measurement unit (IMU) is presented. The filter uses the measurement model of the multi‐state‐constraint Kalman filter. The filter also regularly resets its origin in coordination with the declaration of keyframe images. The keyframe‐to‐keyframe odometry estimates and their covariances are sent to a global back end that represents the global state as a pose graph. The back end is better suited to represent nonlinear uncertainties and incorporate opportunistic global constraints. We also introduce a method to account for front‐end velocity bias in the back‐end optimization. The paper provides simulation and hardware flight‐test results of the front‐end estimator and performs several back‐end optimizations on the front‐end data.