Robust Three-Dimensional Collision Avoidance for Fixed-Wing Unmanned Aerial Systems

Abstract

This work considers a new ”morphing” potential field navigation technique, where potential fields morph to distribute potential cost to areas necessary for collision avoidance in the particular case of fixed-wing unmanned aerial systems, vehicles with high speed and high inertia. The original algorithm, developed for the two-dimensional lateral case, is extended to the third dimension through a layered potential plane approach in both longitudinal and lateral axes. The developed online navigation method is used in series with an integrated guidance and nonlinear model predictive controller, in which robustness is included through the augmentation of frequency dependent weighting matrices in the control structure. The combined approach is demonstrated in full nonlinear six-degrees-of-freedom simulation of a large fixed-wing unmanned aerial system showcasing successful online three-dimensional collision and obstacle avoidance and simultaneous fixed altitude and inclined trajectory following in a congested urban setting while subjected to high crosswind.