Parameter estimation of aircraft dynamics via unscented smoother using multiple data sets

This paper proposes a method for parameter estimation of aircraft dynamics based on multiple sets of test data. It is an extension of a method based on Unscented Rauch-Tung-Striebel Smoother (URTSS) to estimate the state and unknown parameters, combined with Expectation-Maximization (EM) algorithm to estimate unknown statistics, i.e., the mean and covariance of initial states, and the noise covariances. The developed method achieves stable convergence, because non-convex numerical optimization algorithms are not necessary. The effectiveness of the developed method is demonstrated through a parameter estimation problem of a longitudinal motion of an aircraft.