Abstract:

Sequential convex programming (SCP) shows promise for fuel-optimal sparse control of high-thrust satellites in geostationary earth orbit (GEO), but is highly vulnerable to converge to local minima in the neighborhood of an initial guess. In particular, when optimizing for the time at which to perform a maneuver, these algorithms tend to find solutions within a few hours of the times at which they are initialized. In this work, we propose an algorithm that relies on averaged dynamics to form a proxy system with fewer nonconvexities than the true system. We use a consensus-based optimization framework to reach a consensus between the average and the true system, enabling the SCP to explore more of the solution space and enabling larger deviations of the converged solution from the initial guess. We demonstrate the performance of the proposed method against that of standard SCP on a problem in which the goal is to extend the time between east-west station-keeping maneuvers for a GEO satellite.
Simulations are performed using NASA’s General Mission Analysis Tool, a high-fidelity space mission simulator.