Thesis Defense: Algorithms for Generation and Tracking of Fast and Agile Flight Trajectories

High-speed flight through cluttered environments is essential to many time-sensitive robotics applications. It requires motion planning and control algorithms that enable highly accurate maneuvering at the edge of the vehicle's capability. These algorithms must overcome challenges particular to fast and agile flight, such as complex dynamics effects including significant unsteady aerodynamics, and challenging conditions like post-stall and uncoordinated flight. We focus on the development of trajectory generation and tracking algorithms that address these challenges for a quadcopter aircraft, and for a fixed-wing transitioning aircraft that combines vertical take-off and landing (VTOL) with efficient forward flight.

This thesis contains several contributions. We first show that robust control based on incremental nonlinear dynamic inversion (INDI) enables fast and agile flight without depending on a rigorous dynamics model. Based on the INDI technique, we design a comprehensive quadcopter flight control algorithm that achieves accurate trajectory tracking without relying on any vehicle aerodynamics model. Second, we show differential flatness of a global 6DOF flight dynamics model for a tailsitter flying wing transitioning aircraft. We leverage the flat transform to design an INDI flight control algorithm capable of tracking agile aerobatics maneuvers that exploit the entire flight envelope, including sideways knife-edge flight. Third, we present a trajectory generation algorithm that aims to identify the true vehicle capability boundary by efficiently combining analytical, numerical, and experimental evaluations in trajectory optimization. Finally, we demonstrate our contributions in fast and agile flight through elaborate experiments.

Committee: Prof. Sertac Karaman (Chair), Prof. Russ Tedrake (MIT EECS), Prof. Luca Carlone