High-Speed Obstacle-Avoidance with Agile Fixed-Wing Aircraft

Agile fixed-wing aircraft aim to bridge the gap between rotor-craft and conventional fixed-wing aircraft, with the capability of maneuverable and even hovering flight like a rotor-craft, and of efficient long distance flight like a conventional fixed-wing aircraft. Avoiding obstacles in unknown environments is a challenging task with these platforms, as they have complicated dynamics and a limited payload, and they fly at high speeds. In this work, we present an obstacle-avoidance strategy that avoids collisions while steering the aircraft to the goal. The strategy does not rely on a prior map of the environment, or the ability to build a map in real-time, and can be run in real-time on-board the aircraft. We utilize a library of optimal trajectories, both conventional and aerobatic maneuvers, that are solved off-line. A sequence of these trajectories is pieced together to form a collision-free motion plan within the field of view of the depth camera that steers the aircraft towards the goal region. We validate the approach in a high-fidelity simulation environment. The aircraft flies autonomously through a forest-like map to a goal region, using conventional maneuvers such as banked and helical turns, as well as aerobatic maneuvers such as an aggressive turnaround.