Determining the Optimal Aerodynamic Quality of a Re-entry Capsule in the Atmosphere Based on Reinforcement Learning

This paper addresses the critical need for re-entry capsule design in space exploration, focusing on the aerodynamic quality of re-entry capsules. The aerodynamic quality, represented by the lift-to-drag ratio, is pivotal for precise landings. We present an orbital model for re-entry capsules, considering atmospheric conditions and capsule parameters, and derive equations for motion, overload, and heat flux. Employing a deep reinforcement learning strategy, we enable an intelligent agent to optimize aerodynamic mass parameters, refining its policy through interaction with the environment and maximizing a reward function encompassing overload, landing duration, and heat flux. Our approach pioneers the integration of machine learning with traditional spacecraft design, offering a robust framework for future research and optimization in aerospace engineering.