Data Reduction for Optimizing the Attitude Control Dispatch in a Spacecraft

This work originates from a dedicated attitude control dispatch study (discussed in another chapter of this book) carried out in the context of the current NGGM (Next Generation Gravity Mission) ESA (European Space Agency) project. This control dispatch study investigates the optimal thruster layout (in terms of position and orientation) in order to minimize the propellant consumption relevant to the on-board attitude control. The optimization problem in question is based on a time discretization of the mission scenarios to consider. For the specific real-world problem to solve, however, this approach yields a huge set of time intervals (e.g., of the order of 100,000 elements). An ad hoc clustering approach has therefore been conceived with the objective of making the relevant optimization problem computationally treatable. It essentially involves selecting a limited number (e.g., ranging from some hundreds to some thousands) of instants as representative of the whole set of time intervals requested. The present chapter focuses on the specific clustering approach introduced. The selection of the most appropriate clustering criterion and cluster sizing for the specific application to consider is key. In the particular context of this work, the k-means and k-medoids methods are considered together with the Davies–Bouldin index (DBI) and further evaluation criteria.