Attitude model identification of a quadrotor using the subspace identification method

This paper studies attitude model identification of a quadrotor using a new subspace identification method. To begin with a mechanism model, a nonlinear attitude model of the quadrotor is first established, and its linearized model is then obtained under the assumption of an ideal mechanism configuration. The identification framework for the quadrotor is a closed-loop errors-in-variables (EIV) system. In order to estimate the attitude model, the reference attitude input is designed as an M-sequence and the new subspace identification method is applied to estimate a coprime factor (CF) model of the quadrotor from the available reference input and attitude output measurements. The three-axis attitude estimation model can be eventually extracted from the identified CF model. Finally, a numerical simulation is given to verify the accuracy of the estimated attitude models by a redesigned LQR controller.