Modeling and adaptive control of indoor unmanned aerial vehicles

Abstract

The operation of unmanned aerial vehicles (UAVs) in constrained indoor environments presents many unique challenges in control and planning. This thesis investigates modeling, adaptive control and trajectory optimization methods as applied to indoor autonomous flight vehicles in both a theoretical and experimental context. Three types of small-scale UAVs, including a custom-built three-wing tailsitter, are combined with a motion capture system and ground computer network to form a testbed capable of indoor autonomous flight. An L1 adaptive output feedback control design process is presented in which control parameters are systematically determined based on intuitive desired performance and robustness metrics set by the designer. Flight test results using a quadrotor helicopter demonstrate that designer specifications correspond to the expected physical responses. Multi-input multi-output (MIMO) L1 adaptive control is applied to a three-wing tailsitter. An inner-loop body rate adaptation structure is used to bypass the non-linearities of the closed-loop system, producing an adaptive architecture that is invariant to the choice of baseline controller. Simulations and flight experiments confirm that the MIMO adaptive augmentation effectively recovers nominal reference performance of the vehicle in the presence of substantial physical actuator failures. A method for developing a low-fidelity model of propeller-driven UAVs is presented and compared to data collected from flight hardware.

(cont.) The method is used to derive a model of a fixed-wing aerobatic aircraft which is then used by a Gauss pseudospectral optimization tool to find dynamically feasible trajectories for specified flight maneuvers. Several trajectories are generated and implemented on flight hardware to experimentally validate both the modeling and trajectory generation methods.