Tether control using fuzzy reinforcement learning

Abstract

A fuzzy reinforcement learning architecture called GARIC is used to develop a controller for tether control on board the Space Shuttle. The primary objectives were to deploy the Italian satellite weighing 525 kg to a distance of 20 km above the Space Shuttle by means of a conducting tether, to acquire necessary scientific and operational data, and to retrieve the satellite to the shuttle for reuse. Learning experiments were performed during deployment phase where GARIC learned to maintain a tighter dead-band in a small number of trials. The performance of this controller is compared with a controller which uses conventional control theory, and a non-adaptive fuzzy controller. Our results, which were obtained with the Orbital Operations Simulator (OOS) system, demonstrate that more difficult tasks can be learned by a controller based on fuzzy reinforcement learning.<>