Flight controller design, automatic tuning and performance evaluation of quadrotor UAVs

For the last two decades, with the fast development of electronics, cheap and lightweight flight controllers, accelerometers, global positioning system and cameras have become readily available, which lead to a rapid growth of small commercial multi-rotor Unmanned Aerial Vehicles (UAV). The multi-rotor unmanned aerial vehicles are unstable systems due to their physical structures and feedback control systems underpin their operations and are paramount in the design and implementation of unmanned aerial vehicles. This thesis is dedicated to the topics of quadrotor controller design, automatic tuning and performance evaluation of their feedback control systems. Six key contributions are made in this research: (1) critical evaluation of conventional model-based quadrotor UAV controller design; (2) novel quadrotor PID control automatic tuning process with potential wide applications; (3) quadrotor control system performance assessment method via step response identification; (4) design and performance assessment of cascade and centralized quadrotor control using Model Predictive Control (MPC); (5) design and simulation of Quadrotor Discrete-Time One-step-ahead Predictive Control (DOPC) design; (6) design and implementation of two quadrotor UAV test rigs for validations of the proposed control strategies. In this thesis, a critical review of the conventional conventional modelbased quadrotor flight controller design is firstly conducted. Numerical simulations and experimental tests show the simplified integrator model and the neglected dynamics, which are commonly used in the conventional quadrotor controller design, can dramatically affect the closed-loop responses of the systems. The problems existed in the conventional modelbased PID controller design serve the motivation to develop the PID autotuner. The key idea behind the proposed auto-tuner is to fit the dominant dynamics of a physical plant into a simplified integrator plus delay model, which is then used to design the PID controller. The auto-tuner ensures a sound closed-loop control performance, without endangering the unmanned aerial vehicles and providing engineers and practitioners with reliable controller parameters. Another important issue for future quadrotor applications is how to reliably assess its closed-loop control performance and identify potential faults before flight. To address this problem, a novel approach to assess the closed-loop control performance of a quadrotor UAV is proposed in this thesis, which gives an intuitive way to more precisely evaluate the designed controllers and to obtain better controller tuning parameters. The auto-tuner and the proposed performance assessment scheme are evaluated on the self-made test rigs with final validations on flight tests. This thesis also covers the topic of quadrotor linear and nonlinear predictive control. For linear MPC, both centralized MPC and cascade MPC control architectures are designed for the quadrotor position controls. From the comparison, the advantages of cascade MPC over centralized MPC are highlighted. In addition to linear MPC, a novel optimization based discrete-time nonlinear predictive controller (DOPC) is developed. The main benefit of DOPC is to enable the quadrotor to fly with large Euler angles, so that it can make aggressive manurer and handle larger disturbance. Robustness analysis is conducted on the DOPC quadrotor control system