一种基于元强化学习的GEO在轨服务任务规划方法
By employing a meta-reinforcement learning-based approach, the autonomy and efficiency issues of GEO on-orbit servicing mission planning were addressed. This enabled efficient and rapid mission planning for multiple servicing spacecraft across multiple orbital targets, thereby enhancing the autonomy and adaptability of on-orbit servicing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- KUNMING UNIV OF SCI & TECH
- Filing Date
- 2023-10-08
- Publication Date
- 2026-07-17
AI Technical Summary
GEO on-orbit servicing mission planning faces a complex and ever-changing environment and a variety of uncertainties. Existing technologies struggle to achieve rapid, safe, and efficient mission scheduling and autonomous planning, especially when multiple servicing spacecraft are simultaneously serving multiple targets in different orbits, lacking adaptive and cost-maximizing solutions.
A meta-reinforcement learning-based approach is adopted. By inputting initial orbital parameters, the multi-cycle Lambert problem is solved, reinforcement learning parameters for on-orbit service planning tasks are defined, an agent interaction environment is constructed, and initial policies are generated using meta-learning algorithms and deep reinforcement learning algorithms to optimize task planning.
It has achieved autonomy and flexibility in planning autonomous missions for multiple on-orbit servicing spacecraft, and can quickly and effectively solve GEO on-orbit servicing missions, thereby improving the autonomy and adaptability of planning and ensuring the speed and efficiency of missions.
Smart Images
Figure CN117382920B_ABST