Model-Identification of a Quadcopter and Model-Based Design of an Attitude Controller

Abstract

Control system design for a non-linear and complex dynamical system is a challenging task, as model uncertainties result in the degradation of performance. To cater for this challenge, the System Identification (SYSID) approach is being used for the Model-Identification of a system. It involves estimating the system model from its input excitations and output response. Black-box system identification estimates the system model with no prior information known about the system. In this research, the attitude model of a Quadcopter Unmanned Aerial Vehicle (UAV) has been estimated from its input-output data after injecting a designed excitation signal on a specially designed Three Degree of Freedom (3-DOF) platform. Model identification has been done by assuming de-coupled roll, pitch, and yaw rates systems. The estimated state-space model gives up to 90% fit to simulation data, 99% in one-step ahead prediction, and 98% in 3-step ahead prediction. For control design application, the higher order plant model of the model is reduced to 3, dominant region of the root locus is selected while ignoring higher frequencies. Reduced ordered plant error with the full-order plant is acceptable within the bandwidth of the system. The attitude subsystem closed-loop model was built in Simulink using the estimated model. Model validation