ESA GNC Conference Papers Repository

Mission objectives of modern science and Earth observation satellites pose stringent performance requirements on the on-board attitude and orbit control system (AOCS). The AOCS performance strongly depends on the knowledge of the spacecraft dynamics involved in the control loop. Most of the commonly used control system architectures make use of feed-forward and feed-back paths. The feed-forward path is used in order to improve the system response on the commands, whereas the feedback path is used to compensate for various perturbations within the control path. Among different kinds of perturbations acting on the spacecraft during in-orbit operation, the modelling uncertainty is the most dominant one, especially when satellites with high agility are considered. In consequence, improvement of the accuracy of the feed-forward path minimizes the burden of the feed-back controller, and hence improves the overall responsive performance of the system. Most of the feed-forward controllers are based on the torque calculation to be imparted to the spacecraft in order to achieve some guided reference acceleration and depend on the moment of inertia tensor of the spacecraft. The moment of inertia is typically estimated on-ground using various analytical and experimental techniques; however it may deviate during in-orbit operation as a result of fuel depletion, anomalous performance characterizing un-deployed or broken appendages, or by docking to an object with unknown mass properties. This paper addresses determination of moment of inertia characteristics of a spacecraft using in-orbit data. The method is based on system identification methodology that includes a tunable plant model reflecting the spacecraft dynamics. It is subject to the measured inputs from various on-board sensors to minimize the identification error, i.e. the difference between measured and modelled quantities. Various identification approaches involving offline and online processing methods are presented and discussed. The offline parameter estimation is performed in post-processing mode and allows a superior treatment of the measurements and a superior physical description of the involved dynamics, including nonlinear dynamics. On the other hand, the online parameter estimation is performed in real-time and allows parameter estimation on-board. The latter is of particular importance when rapidly changing moment of inertia parameters need to be determined. Finally, results based on experimental data obtained from the ground-based satellite dynamics hardware demonstrator are shown.