Pseudo-Random Binary Series Based Comparative Study on State Estimation for a Quadrotor

Abstract

Quadcopter has many uses in everyday life. So it requires advanced controller to control its attitude. Due to their good applicability to nonlinear multivariable complex control objectives, numerous advanced state-feedback controller design methods have been applied in the quadrotor attitude control. State estimator plays great role in controller design in which state of a plant is feedback to input side. However, traditional observer have reduced performance due to input noise and ouput noise. This thesis proposes a subspace identification-based state estimation method suited to the state-variable estimation for the quadrotor attitude control system in the absence of models with disturbing noises. Above all, an observer-based discrete state feedback control system is designed starting from attitude control performances given by users. In order to obtain a high accuracy model and state estimation, a sufficiently excited attitude-angle reference input is designed as a pseudo-random binary series. Then, the control quantity and the real attitude angle being input and an output for the quadcopter, respectively, can be collected as outputs. The existing general attitude model for the yaw channel is used to test the validation of data. Next, this thesis proposes a 4SID-based state estimation method to deal with the yaw-axis. Quadcopter yaw angle attitude state estimation problem under the conditions that the measurements for the control quantity and the output attitude are corrupted with disturbing noises and that there are no available models. Based on identified model for the generalized attitude model from the proposed state estimation method, the Kalman filter and smoother are applicable and estimation-accuracy comparison done. Finally, the simulation results showed that among the stated estimtors, a 4SID based state estimation method is the best estimator.