Nonlinear Recursive Estimation of Boost Trajectories, Including Batch Initialization and Burnout Estimation

Abstract

A 12-state, iterated nonlinear recursive filter is formulated for boost trajectory estimation and prediction. The recursive filter is initialized with state and covariance estimates from a polynomial batch filter. An interacting multiple-model technique handles staging and burnout events by statistically blending boost filter estimates with estimates from a Keplerian filter. A new contribution is a boost prediction model that includes terms that model the nonlinear growth of thrust acceleration magnitude caused by propellant consumption and variations in thrust orientation associated with gravity-turn maneuvers and steering maneuvers at constant or variable angles of attack. State estimates and covariances are iterated at each filter update, subject to constraints on magnitude and orientation of the thrust acceleration estimates. These constraints are useful in controlling unrealistic estimates caused by poor measurements, in precluding unbounded acceleration predictions during long intervals without measurements, and in handling acceleration discontinuities at staging events. A noteworthy feature is that this algorithm does not rely on any a priori information such as a booster template.