On the Application of Reinforcement Learning in Multi-debris Active Removal Mission Planning

Abstract

We formulate the active multi-debris removal mission planning task as a Reinforcement-Learning (RL) problem and developed an adjusted Deep Q-Learning (DQN) solution. We propose novel definitions of the state space, action sets, and rewards in the context of active multi-debris removal mission planning. These definitions facilitate recasting the mission planning problem into a RL problem. As such, a powerful DQN algorithm may be applied to solve the mission planning problem using an RL approach. We test this new approach using a subset of Iridium 33 debris cloud. Very encouraging results are observed. Future applications to a reactive autonomous space mission planner are also discussed.