关键词: 小天体探测, 柔性附着, 着陆制导, 智能制导, 强化学习

Abstract: A flexible lander constructed with compliant material can effectively reduce the risks of overturning and rebound during landing in weak-gravity environments. However, the strong nonlinear dynamics introduced by the flexible structure and the requirement for multi-node cooperation increase the difficulty of guidance design. To address the cooperative guidance problem for flexible landing, an adaptive intelligent guidance method is proposed. First, a simplified trajectory planning model for flexible landing is developed, enabling efficient generation of reference trajectories while preserving the primary motion characteristics of the system. Furthermore, the stable trackability of the reference trajectory by the flexible lander is established, based on which a reinforcement learning-assisted guidance law is constructed. Within this framework, reinforcement learning is only used for online parameter adaptation and compensating-input generation, thereby reducing the learning complexity. By incorporating performance indices such as attitude regulation and constraint satisfaction into the reward function, the learned policy is able to improve landing steadiness and reduce fuel consumption while maintaining accurate tracking of the reference trajectory. Finally, cooperative guidance simulations for flexible landing are conducted based on the small celestial body 433 Eros, demonstrating the effectiveness of the proposed adaptive guidance method.

Key words: small celestial body exploration, flexible landing, landing guidance, intelligent guidance, reinforcement learning