Improved Reinforcement Learning–Based Method for Emergency Mission Planning to Remove Space Debris

Abstract

A mission planning method with an online optimization mechanism based on improved reinforcement learning was proposed to address the task planning problem of space debris removal tools utilizing stochastic maneuvering strategy and target selection strategy in active space debris removal scenarios. The primary objective of this method is to enhance the efficiency of space debris removal operations within a specified timeframe. The initial steps involve establishing working models for lasers and onorbit working vehicles, as well as formulating an optimized reinforcement learning improvement approach. Subsequently, a reinforcement learning training module dedicated to space debris removal tasks was introduced. This module focuses on training the maneuver location and work objectives of the two space debris removal tools while learning the optimal strategy for achieving a successful removal rate. The trained decision-making strategy is then integrated into an optimizer to analyze the cost consumption associated with the planning outcomes. The key finding of this study is the development of a stable space debris removal strategy through the training of the removal tool. By choosing an appropriate agent, the agent-based optimization process achieves a 100% space debris removal rate, leading to cost savings in ground-based movement of removal tools and improved mission planning.