Spacecraft Onboard Mission Planning for Collision Avoidance via Imitation Learning

The increasing amount of space debris has significantly raised the risk of orbital collisions, requiring rapid and reliable autonomous threat avoidance strategies for spacecraft. Traditional planning methods lack the ability to handle complex environments and fail to incorporate expert knowledge effectively. This paper introduces a hypergraph-based state representation that models the spacecraft's operational status and actions. Subsequently, a heuristic learning algorithm is proposed to incorporate domain expertise extracted from optimal plans. The simulation results indicate that this method completes the spacecraft autonomous mission planning in different threat avoidance scenarios within a short time. The proposed approach offers an effective and scalable solution for autonomous spacecraft collision avoidance mission planning.