Comparison of the Extended Kalman Filter and Maximum Likelihood Methods for Parameter Identification of an Autonomously Piloted Vehicle

Abstract

A six-degree-of-freedom airframe model for an autonomously piloted vehicle (APV) is first constructed containing a number of aerodynamic coefficients. The coefficients model include twenty four parameters. The extended Kaiman filter (EKF) and maximum likelihood (ML) methods are introduced to identifying the aerodynamic coefficients and applied to the APV. The two methods are compared by simulated flight test data. The results show that when initial guesses are poor, the ML method provides a better result. On the contrary, if initial guesses are close to the true values, the iterative EKF method is superior to the ML. Thus, the paper demonstrates that the EKF and/or ML method may be employed to improve the aircraft modeling by processing postflight data.