Stability Derivative Identification Using Adaptive Robust Extended Kalman Filter for Multirotor Unmanned Aerial Vehicle (M-UAV)

The invention of Unmanned Aerial Vehicle (UAV) in the early 1900 for military purposes and UAV applications in commercial purposes afterwards in the early 2000 accelerate its research in many engineering fields. Multirotor UAV such as quadrotor is usually unstable without a flight controller. Consequently, a proper and accurate model of the UAV dynamics is essential for its system stability. System identification allows the researchers to create an accurate parameter to the mathematical model of a dynamic system based on measured data. This work emphasizes in designing a robust and adaptive filter to develop an accurate mathematical model based on Newton-Euler method which includes aerodynamic drag and moment which are necessary in determining the correct model prediction. The focus of the present work is mainly on Kalman filter development for parameter estimation. While Kalman filter is only efficient in linear problem, an extended version of the filter itself deals with the nonlinear problem in which most real problem is actually nonlinear. This work investigates the performance of the extended version of the Kalman filter relative to parameter estimation. The performance of the filters are evaluated based on their estimation with the actual recorded flight data and presented based on data overlapping using Root Mean Square Error (RMSE). Ardupilot APM is used to acquire the actual flight test data and MATLAB is utilized to carry out the state estimation. To evaluate the performances of the filters, Goodness of Fit (GOF) approach was used. It is found that the GOF index of the present approach is 0.853 which is 25% higher than that of the Robust Extended Kalman Filter approach for the present flight test result.