Real-time parameter estimation of unmanned aircraft dynamic models provides valuable information about the physical model of a vehicle. However, the aircraft model estimation performance can be severely degraded with an active control system and high input collinearity such as those found on a quadrotor unmanned aircraft. Orthogonal multisine input signals are superimposed with the active control system and utilized for real-time recursive Fourier Transform Regression parameter estimation. Experiments were conducted with varying input signal amplitudes and regressor model structures. Each configuration was evaluated on a quadrotor with nominal hardware conditions and with a damaged propeller. Flight test generated parameter estimates were compared with benchmark motor/propeller values. The results indicate as the multisine amplitude increases, the iden-tifiabllity of individual actuator effectiveness increases with a model structure containing each actuator. However, the increasing multisine amplitude induces unwanted rotational perturbations. Roll and pitch model parameter estimates can be quickly and accurately estimated with the estimate methodology with approximately double the perturbations compared with solely the flight controller. Overall, a well designed coupled multisine input signal and Fourier Transform regression provides an efficient, accurate real-time parameter estimation technique that can effectively overcome challenges imposed by closed loop system identification systems with high input collinearity.
展开▼