A Novel Brain-inspired Architecture and Flight Experiments for Autonomous Maneuvering Flight of Unmanned Aerial Vehicles

Autonomous maneuvering flight is essential for civil applications of unmanned aerial vehicles(UAVs). Yet it remains a challenge due to the coupled translational and rotational dynamics and the complexity of the movements. This paper proposes a brain-inspired architecture for autonomous maneuvering flight of miniature UAVs. In designing such system, several key issues will be addressed: maneuver controller design, online maneuver decision making, and command generation. First, a modeling method based on dual quaternion is proposed to describe maneuvering flight. Then, an incremental flight control method in 6-DOF is introduced. Furthermore, a novel dual quaternion-based dynamic maneuvering primitive(DQ-DMP) is proposed to learn the trajectories performed by human pilots. Finally, a decision-making method based on a spiking neural network and Actor-Critic network is provided for the choice of maneuvering primitives. The hardware-in-the-loop simulation and outdoor flight test show that the architecture can learn, repeat and generalize the maneuvers. Various aerobatic maneuvers can be performed by the fixed-wing UAV after several demonstrations.