Date of Award
Master of Science in Electrical Engineering (MSEE)
Dr. Addy Wahyudie
This research intended to design, analyze, and implement a robust controller for a quad rotor system and compare the designed robust controller with a proportional-integral-derivative (PID) controller. The ArduCopter platform was used as a target system with a 3DR airframe and necessary modifications, a system model, and system identification processes were executed as prerequisite steps to reach the objective.
The work in this thesis includes exploring the existing research on this topic and builds on the results presented in these previous studies to add value to the scope of quad-rotor system control. During this study, system modeling was conducted, where a near-hover non-linear model for the system was obtained and realized in Simulink. Furthermore, system identification was performed to obtain the platform parameters, which include the blade thrust coefficient, inertias, and propeller drag coefficient. The identification process was based on standalone experiments as well as flight data and the non-linear model was validated and assured to be representative of the real system. The control system was then designed with both classical PID and robust controllers. This control architecture was designed to be scalable to other platforms. The classical controller was designed analytically for the body rate loop, while root-locus plots were then used for the attitude loop. The robust controller was designed based on the H∞ method and the augmented plant was constructed using the GS/T scheme. The existing software for the ArduCopter was modified to have a customized logging structure and flight modes functionality, and to make it suitable to implement a robust controller in state-space form. Finally, experimental flights were conducted to tune the classical controller and test the robust controller, and to conduct robustness tests by injecting user-controlled, known disturbances in flight.
Various outcomes were reached and findings were made along the research stages. One of the outcomes reached, was to determine the effectiveness of the identification methods used, despite the shortages in the standalone experimental setup. Furthermore, the yaw torque model reused from previous studies was found to not match properly with the flight data. The drag on propeller rotation as presented in the literature is considered to be dominant over the anti-torque action. The flight data and analysis thereof in this research show that anti-torque contributes more to generating yaw torque than propeller drag.
Furthermore, a comparison between the PD controller and the robust controller was made during the experimental flights. The flight data showed that the PD controller has a good dynamical response, but lacks robustness against imbalance in actuation (or untrimmed actuators). Integrator action was added gradually over the course of a few experiments to enhance the performance without affecting the dynamical response. The tuned controller showed fairly good overall robustness when disturbances were injected manually. In comparison with the PD controller, the robust controller performed far better in terms of dynamical response and disturbances rejection, but the controller obtained is much more complex than PID controller and requires more computational time to propagate over time.
Al-Ali, Ismail Sulaiman, "Parameter Identification and Robust Control Applied to a Quadrotor" (2014). Theses. 517.