Flight Vehicle System Identification via Nonlinear Smoothing Combined with Unscented Transform

This paper describes a new algorithm for system identification of flight vehicles that is expressed by nonlinear state and observation equations with unknown parameters. The developed algorithm employs a nonlinear smoothing technique based on nonlinear programming with an unscented transform. It covers general classes of problems, including those which have non-additive process noise, unknown covariances of the process noise and measurement noise. By applying the developed algorithm to the problems of identifying the ballistic coefficient of a reentry body and the aerodynamic coefficients of a research aircraft, good identification accuracies comparable to those of established methods as well as a stable convergence property is demonstrated.