Design and Implementation of an on-board control system for quadrotors

This thesis presents the design, implementation, and validation of an Incremental Nonlinear Dynamic Inversion (INDI) controller for a quadrotor using the PX4 framework, following a V-shaped model validation approach. The primary objective of this research is to develop a robust controller capable of executing autonomous missions under model uncertainties and external disturbances, while also assessing its fault-tolerant capabilities and response to aggressive manoeuvres. The INDI controller is designed in combination with a proportional-derivative (PD) controller to enhance stability and robustness. The validation process begins with Model-in-the-Loop (MIL) simulations, conducted in MATLAB and Simulink, to establish an initial tuning of the control parameters based on a high-fidelity quadrotor model. These simulations provide a first insight into the expected closed-loop response and dynamic behavior of the system. Following MIL validation, Software-in-the-Loop (SIL) simulations are performed to ensure compatibility between the generated C++ code and the PX4 autopilot software architecture. At this stage, the controller is integrated into the PX4 firmware, and its performance is evaluated within the simulation environment, verifying that the numerical implementation remains consistent with the initial theoretical formulation. The next phase involves Processor-in-the-Loop (PIL) simulations, where the controller is executed on the CubePilot Black FMUv3 flight controller to assess real-time computational feasibility and onboard execution performance. This step is crucial to ensure that the embedded system can handle the control law in real-world scenarios without excessive delays or computational bottlenecks.\\ The controller’s performance is evaluated through a series of increasingly complex trajectory tracking experiments, including circular trajectories, figure-eight patterns, and waypoint-based navigation tasks. The experimental results are analyzed to quantify the controller's accuracy, robustness, and ability to reject disturbances. Furthermore, additional test scenarios explore the response variations in the system’s mass and inertia. This research not only demonstrates the feasibility of implementing advanced nonlinear control strategies on resource-constrained embedded flight controllers but also provides a structured methodology for bridging the gap between theoretical control design and real-world deployment in aerial robotics.