Safe Capturing and Stabilizing of Diverse Debris Objects

Abstract

With increased access to space, the challenge of space debris is becoming a more and more relevant topic. To keep space usable for future generations, space debris has to be mitigated. The most effective mitigation measure includes Active Debris Removal (ADR) of around 5 objects per year. This thesis investigates the challenge of the Attitude Control System (ACS) in ADR. The problem here is that often the exact size, shape and mass of the targeted debris objects are unknown. Furthermore, through the capturing process the mass and inertia of the satellite system is changed, making it difficult for the ACS to maintain a stable position. In this thesis an approach of Deep Reinforcement Learning (DRL) algorithm is investigated to solve this satellite attitude problem. A DRL approach will help in coping with the unknown inertia and will enable the ADR to handle diverse debris objects. Several different attitude control scenarios are investigated, with the complexity increasing continuously.