In recent years, quadrotors have garnered significant attention in both industry and academia due to their excellent maneuverability and hovering ability. This results from onboard sensors with high accuracy, remote controllers with high performance, and network communication among sensors, controllers, and plants with high efficiency. A quadrotor control system of this type can be regarded as a networked control system (NCS) enjoying remarkable scalability, resource efficiency, and ease of maintenance. However, new challenges in controller design arise from network-induced issues. Model predictive control (MPC), as an optimization-based control method, is able to provide not only the optimal control input for the current time instant but also the predicted state and input sequences, which provide a promising solution to handle network-induced issues. Moreover, the state and input constraints that exist in many applications can be effectively dealt with by MPC. These appealing features have motivated the development of many MPC schemes for quadrotors and NCSs. However, how to effectively solve network-related problems by MPC, and how different factors in MPC implementation affect the control performance are still open problems.

We propose a robust output feedback MPC framework for constrained networked quadrotor control systems subject to packet dropouts and external disturbances. The packet dropouts randomly happen in both sensor-controller (S-C) and controller-actuator (C-A) channels. The proposed output feedback MPC scheme consists of a state observer that accommodates the random measurement loss and a state feedback MPC that stabilizes the perturbed system. The proposed observer enables the estimation error dynamics to be represented by a switched system. By developing a generalized robust positive invariant (GRPI) set under the switched system formulation, the estimation error can be confined to this invariant set, which serves as the explicit error bound of state estimation. Similarly, an extended robust positive invariant (ERPI) set is developed to describe all possible realizations of the deviation between the predicted and actual state. Then, the GRPI and ERPI sets are utilized to tighten the state and input constraints to alleviate the effects of random packet dropouts and disturbances. By imposing tightened constraints on the predicted states and inputs in the optimal control problem, the system can be stabilized by the proposed output feedback MPC scheme with guaranteed constraint satisfaction. Simulation results are provided to validate the effectiveness of the proposed method.

Three MPC schemes are adopted and compared for quadrotor control, including conventional MPC, tube-based MPC, and Lyapunov MPC. Moreover, different factors that may affect the control performance are considered in a dual-loop control framework. Firstly, since the disturbances usually appear in practical implementations, the robustness of three MPC schemes against different levels of disturbances is evaluated and compared. Then, to simulate the real control processes and validate the effectiveness of three MPC frameworks, the control inputs generated by three controllers with different prediction models are applied to the same nonlinear quadrotor system. Moreover, since the sampling rates of inner and outer control loops in the dual-loop control framework are usually assumed to be the same, we explore how different dual-loop sampling ratios affect the control performance, which facilitates the controller design for quadrotors and provides a direction for theoretical studies.

Finally, after concluding the obtained results, future study directions in quadrotor control are provided at the end of this thesis.