High-Performance Flight Control of Variable-Pitch-Propeller Quadcopters

Variable-pitch-propeller (VPP) quadcopters are a new type of micro aerial vehicle. Compared to conventional fixed-pitch-propeller quadcopters, VPP quadcopters can not only generate positive but also negative thrust and hence fly upside down. They also outperform the conventional quadcopter in terms of control bandwidth and manoeuvrability. Despite the advantages of VPP quadcopters, their potential has not been well explored yet, though there exist some relevant studies in the literature. The aim of this thesis is to explore the unique features of VPP quadcopters in two important aspects. The first aspect is about flight safety: that is, how to ensure a VPP quadcopter flies safely even when a propeller is partially damaged. Two fault-tolerant controllers for different types of VPP quadcopters are proposed, respectively. It is found that fault-tolerant control of a VPP quadcopter exhibits unique and interesting features compared to conventional quadcopters. In particular, a separated-powered VPP quadcopter is fully controllable with only three actuators. With the proposed control law, a faulty centralized-powered VPP quadcopter can track the reference path. The system and simulation experiments are established and conducted in a physical driven environment called SimScape, to verify the effectiveness of the proposed control laws. Real experiments are conducted to verify the dynamic model of VPP. The second aspect is about flight agility: that is how to fully utilize the ability of VPP quadcopters to achieve highly agile flight. In particular, novel control laws are proposed that can successfully achieve the tic-toc flight task, which is one of the most challenging manoeuvrers of aerial vehicles. Thanks to the feature that a VPP could generate either positive or negative thrusts, such a task is achievable by VPP quadcopters. The tic-toc flight is challenging to achieve partially because it is far from the hovering equilibrium state. A state-of-the-art reinforcement learning algorithm is employed to solve this problem. An extension algorithm for performance adjustment and controller redeployment is proposed as well. Simulation results verify the effectiveness of our control laws.