| Investigating the Hyperparameters of Reinforcement Learning Effects on Orbit Raising Maneuver |
| کد مقاله : 1156-AERO2024 |
| نویسندگان |
|
مجید بختیاری *1، حامد سلیمانی2، کامران دانشجو1 1عضو هیئت علمی دانشگاه علم و صنعت ایران 2دانشجوی دکترا رشته مهندسی هوافضا دانشگاه علم و صنعت ایران |
| چکیده مقاله |
| In recent years, significant advancements in the field of artificial intelligence have prompted space research, particularly in orbital missions, to increasingly embrace these methods, with a specific focus on machine learning. In this research, considering the dynamics of circular in-plane low-thrust orbit transfer based on the equinoctial differential equations as the environment for establishing agent interaction, a continuous space for the problem variables which are the six equinoctial orbital elements of a spacecraft, a model-free algorithm called Actor-Critic algorithm, is implemented. The action space which defined as a thrust vector is applied to the environment under a policy, and the agent is trained by Actor-Critic algorithm, to be capable of performing the LEO to GEO low-thrust transfer. Effects of the hyperparameters such as the discount factor, learning rate and the number of nodes in actor and critic network, are investigated in this scenario. It is shown that increasing the discount factor and learning rate assists the trained agent in operating accurately in the environment of the orbital transfer problem. Increase in the number of nodes in the neural network cause an increment in the learning time of the agent. By increasing amount of the discount factor near to 1, the agent performs some further searches in the environment to find other possible optimal policies. After two training processes, one can use the trained agent in different cases with similar dynamics to the main problem, and there is no need to adjust or re-simulate the parameters and dynamics of the problem. |
| کلیدواژه ها |
| Low-Thrust – Equinoctial orbital elements – Reinforcement Learning – Actor-Critic networks – Agent |
| وضعیت: پذیرفته شده برای ارائه شفاهی |
