首页
>
期刊首页
>
系统工程与电子技术（英文版）
>
2026年4期
>
A new data-driven task planning approach for on-orbit refueling of low-earth orbit mega-constellation members
A new data-driven task planning approach for on-orbit refueling of low-earth orbit mega-constellation members
摘要
On-orbit refueling for low-Earth orbit(LEO)mega-con-stellations requires solving complex access sequence problems.The increasing number of satellites complicates finding optimal solutions.To tackle this,we introduce a data-driven enhanced ant colony optimization(DDEACO)method in this paper.DDEACO uses a two-step approach:initially,it trains a deep Q-network(DQN)using the ∆V transfer cost as a reward and employs gated recurrent units(GRUs)to learn Q values for selecting the optimal next satellite member for visitation,captur-ing the sequences'temporal dependencies.Then,it integrates the pre-trained GRU with traditional ant colony optimization(ACO)to offer new heuristic paths,improving ACO's exploration and efficiency.The effectiveness of DDEACO is demonstrated through numerical experiments on systems tool kit(STK)-gener-ated datasets with 30 satellites and 200 satellites,showing its significant performance.
年，卷(期)：
2026, 37(4)
机标分类号：
A811
TN929.5
TP27
在线出版日期：
2026年9月10日
