Reinforcement Learning-Based Solar Sail Trajectory Design

Solar sailing is an emerging technology that uses solar radiation pressure to propel spacecraft, allowing longer-duration and higher Delta- V missions. However, space trajectories, including solar sail trajectories, are subject to uncertainty due to navigation errors. It is important to properly account for this uncertainty and its associated risks in the trajectory design process. Building upon the recent development of deep reinforcement learning (DRL) and its successful application in spacecraft guidance dynamics and control, we propose a new formulation for DRL-based solar sail trajectory optimization under uncertainty with practical mission considerations (specifically, imperfect SRP reflection and sail attitude constraints). The objective of this optimization is to minimize the solar angle (the cone angle between the sail normal and sunlight vectors) across the trajectory, which is crucial for missions that require fixed times of flight, such as scientific observations with specific lighting conditions or gravity assistance at specific epochs. Monte Carlo simulations demonstrate the advantages of DRL-based robust solar sail trajectory optimization with the minimum solar angle objective, which can generate a nominal trajectory and a corresponding closed-loop guidance law simultaneously.