Recursive Multiplicative Estimation of Rigid Body Attitude and Moment of Inertia

Typically, the center of mass and inertia tensor of a spacecraft are estimated pre-launch using a mass properties table and/or a CAD model of the spacecraft. If these estimates are not sufficiently accurate to meet mission requirements, they can be refined after launch using sensor data from the spacecraft. In order to refine pre-launch estimates of these properties, variations of the Kalman filter are developed to estimate the attitude and angular rate of the spacecraft, as well as a representation of the inertia tensor. Two parametrizations of the inertia tensor are investigated: the standard moments and products of inertia representation and a representation formed from the principal moments of inertia and the attitude of the principal axes. Monte Carlo analysis is considered for three scenarios in which the mass properties are estimated, and it is shown that both methods are statistically consistent and capable of improving upon inertia estimates but suffer from linearization error when uncertainty in the moment of inertia tensor is large