Deep Reinforcement Learning Model for High Level Science Planning using Formation Flying(NPO-51314-1)

The software trains algorithms to optimize science return for Earth orbiting missions that use formation flight. The usage of deep learning techniques is uncommon for ground planning systems and has never been used before for onboard systems. The software has been designed with future onboard usage in mind. The software creates a plan for each satellite in the formation to maximize science data return and minimize fuel consumption.