Closed-loop control of spacecraft formations with applications on SPHERES

Abstract

Formation flying satellites offer potentially greater science returns and operational capabilities than attainable with a monolithic spacecraft. Successful control of a formation of spacecraft can be divided into two separate stages. The first stage creates a plan that meets a set of mission objectives, and the second stage implements the plan. Plans are specified as a sequence of [delta]V commands executed at specific times during an orbit. This thesis presents an online method for generating fleet-wide plans, using convex optimization techniques, that satisfy multiple objectives. The approach allows for minimum and balanced fuel usage, can position spacecraft in arbitrary configurations, and favors low-maintenance orbits that do not drift apart. Additionally, the architecture is applicable not only to formation-keeping maneuvers, but also to formation reconfigurations. Various simulations demonstrate the importance of accurately implementing plans for formation flying as well as autonomous rendezvous and docking missions. Specifically, the relationships between process error, overall fuel use, and position error are studied. Theory is put into practice with the development of a new low-level, closed-loop thrust controller for the Synchronized Position Hold Engage and Reorient Experimental Satellites (SPHERES). The controller processes measurements from accelerometers and gyroscopes to monitor thruster performance in real-time. Experiments conducted on the International Space Station (ISS) validate the controller and establish a foundation for future enhancements to the underlying algorithm. Finally, data from a series of high-fidelity formation flying simulations is presented that confirms the analysis done elsewhere in the thesis.

(cont.) The multi-objective planner is used in a closed-loop control system that guides a formation of five spacecraft through a hypothetical mission involving both reconfigurations and formation-keeping. Data from the simulations allows a straightforward, side by side comparison of the effects and relative importance of sensor error versus implementation error.