UNMANNED AERIAL VEHICLE DYNAMIC MODEL IDENTIFICATION USING MIXED ESTIMATION BEFORE AND AFTER MODELING METHOD

Abstract

The unmanned aerial vehicle dynamic model is identified with the use of combination of two different methods called Estimation before Modeling and Estimation after Modeling, using Kalman filter and available measurements. Tuning of filter is one of the difficult stages of the estimation using Kalman filter. It can be made easier to tune Kalman filter by using the Estimation before Modeling method because in this method estimation and modeling of aerodynamic forces and moments are done in two stages. In the Estimation before Modeling method at the first-stage aerodynamic forces and moments are estimated without a priori structure and in the second stage estimated forces and moments are modeled versus suitable state variables. Dimension of augmented state vector for Kalman filter is reduced by using the Estimation before Modeling method. Combination of the Estimation before Modeling and Estimation after Modeling methods are suggested to achieve observability and simplicity of filter tuning. Linear accelerometers out of the center of gravity purposely are used to measure the linear and angular accelerations to achieve better observability. Bounds of uncertainties for estimated aerodynamic coefficients are calculated using diagonal elements of covariance matrix. Suggested method can be used to identify any flight vehicle dynamics and find accurate mathematical models for flight simulators.