Deep-Learning-Based Optimal Relative Trajectory Design for Space Environment Governance

Abstract As human space activities rapidly advance, the increasing amount of space debris endangers the space environment and necessitates the autonomous active debris removal (ADR) to ensure the sustainability of space resources and safety of future missions. In the ADR process, a chaser first approaches a piece of debris via long-range transfer, followed by close-range maneuvers that facilitate target characterization, inspection, and capture. A key challenge in ADR is the close-range trajectory planning, which resembles formation reconfiguration. However, the autonomous ADR complicates the design of close-range rendezvous trajectories due to the need for repeated orbit propagation and the substantial computational burden, which limits the chaser’s responsiveness. To address the challenges in space environment governance, this paper presents a rapid trajectory design approach for close-range rendezvous during autonomous ADR, utilizing optimal thrust based on deep learning techniques. By leveraging deep neural network, the computational efficiency and responsiveness of trajectory design in space environment governance are significantly enhanced.