System identification of a dron mounted on a gyroscopic test bench

This dissertation applies an experimental methodology to study and identify a quadrotor mounted on a gyroscopic test bench. Its aim is to find suitable data-driven linear timeinvariant (LTI) models in both continuous and discrete time for pitching, rolling and yawing motions separately and to valida...

Descripción completa

Detalles Bibliográficos
Autor: Di Giuseppe, Gabriel Adrian
Tipo de recurso: tesis de maestría
Fecha de publicación:2023
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/398283
Acceso en línea:https://hdl.handle.net/2117/398283
Access Level:acceso abierto
Palabra clave:Helicopters
Gyroscopes
Drone aircraft
System identification
Control
Quadrotor
Gyroscope
Drones
Test bench
Helicòpters
Giroscopis
Avions no tripulats
Àrees temàtiques de la UPC::Informàtica::Automàtica i control
Descripción
Sumario:This dissertation applies an experimental methodology to study and identify a quadrotor mounted on a gyroscopic test bench. Its aim is to find suitable data-driven linear timeinvariant (LTI) models in both continuous and discrete time for pitching, rolling and yawing motions separately and to validate them. A literature review is done to assess the significance of UAVs in general, the interest in them, and their impact. The work also studies the current methodologies utilized in research to study, identify and control UAVs. Prior documentation and knowledge of the chosen drone is assessed to understand the building foundations of the dissertation. An experimental, drone-independent methodology is developed for the work. New userfriendly interfaces and tools for experimentation and control of the drones at UPC ESEIAAT are developed in the MathWorks’ framework, which are independent of the selected approach. Guidelines for them are written for future use. In accordance with it, basic physical parameters (i.e., friction coefficient and inertia) are estimated based on an equation that explains the motion of the drone-gyroscope system without the actuations of the motors. Thereafter, data is collected in a system identification framework, and model structures are chosen. The models are then estimated for pitch, roll and yaw separately. Based on them, Proportional-Integral-Derivative (PID) controllers are tuned using the pole placement method and iterative algorithms. Moreover, the models are successfully validated in the platform with the help of the attitude controllers. The platform demonstrates to be a useful tool for both educational and research purposes. It allows the implementation of the proposed methodology in a controlled environment. Also, the system shows signs of nonlinear behaviour, such as nonlinear proportional gains and dead zones. The first two angles linear dynamics were identified to be of second order, while the yaw proved to have a natural integrator, together with a first order dynamic. In accordance with the influence of the gyroscope’s inertia on the drone, the roll was the fastest, and the pitch was in a middle point. Both proved to have one equilibrium point, while the yaw had an infinite amount. Additionally, the rotational friction was pinpointed as the cause of its dead zone, which provoked an offset during the validation phase