Decision-Making for Satellite Anti-Interception Missions Leveraging Multi-Agent Reinforcement Learning

Many-to-many spacecraft autonomous evasion missions necessitate highly coordinated decision-making among several spacecraft to successfully avoid interceptions. Conventional control methods often struggle to manage the inherent complexity and uncertainty of these missions. In response to this challe...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2023 5th International Conference on Data-driven Optimization of Complex Systems (DOCS) s. 1 - 7
Hlavní autoři: Chen, Zixuan, Wang, Jianqi, Wang, Dan, Yu, Sheng, Huo, Jing, Gao, Yang
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 22.09.2023
Témata:
Heuristic algorithms
Machine learning algorithms
multi-agent reinforcement learning
Reinforcement learning
Satellites
Space vehicles
spacecraft evasion missions
spacecraft systems
Training
Uncertainty
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Many-to-many spacecraft autonomous evasion missions necessitate highly coordinated decision-making among several spacecraft to successfully avoid interceptions. Conventional control methods often struggle to manage the inherent complexity and uncertainty of these missions. In response to this challenge, we employ multi-agent reinforcement learning (MARL) algorithms that draw upon machine learning and game theory concepts. Our work is devoted to implementing MARL to facilitate autonomous and intelligent evasion maneuvers by spacecraft, considering the dynamic nature of the space environment and the multiple agent interactions. We utilize the Satellite Tool Kit (STK) as a simulation environment and assess cutting-edge MARL algorithms, with the goal of showcasing the potential of MARL in complex spacecraft evasion missions. Our research endeavors to enhance the autonomy, adaptability, and mission success rate of spacecraft systems under unpredictable circumstances, thereby facilitating more intelligent and adaptive space exploration.
DOI:10.1109/DOCS60977.2023.10294901