摘要
为应对多星环境中复杂多约束条件下的任务分配场景,提出一种多星自主决策观测任务分配算法,该算法采用基于集中式训练、分布式执行的多智能体深度强化学习算法。通过这种方式训练后的卫星智能体,即使在没有中心决策节点或通信受限的情况下,仍具有一定的自主协同能力及独立实现多星观测任务的高效分配能力。
To address the task allocation scenario under complex and constrained conditions in a multi-satellite environment,a multi-satellite autonomous decision-making observation task allocation algorithm is proposed The algorithm uses a multi-agent deep reinforcement learning algorithm based on centralized training and distributed execution The satellite agents trained by this algorithm have certain autonomous collaboration capabilities and the ability to independently achieve the efficient allocation of multi-satellite observation tasks even if there is no central decision-making node or communication restriction.
作者
王桢朗
何慧群
周军
金云飞
WANG Zhenlang;HE Huiqun;ZHOU Jun;JIN Yunfei(Shanghai Satellite Engineering Institute,Shanghai 201109,China;Shanghai Academy of Spaceflight Technology,Shanghai 201109,China)
出处
《上海航天(中英文)》
CSCD
2024年第1期108-115,共8页
Aerospace Shanghai(Chinese&English)
关键词
多智能体系统
深度强化学习
多星系统
多智能体深度确定性策略梯度算法
任务规划
multi-agent system
deep reinforcement learning
multi-satellite system
multi-agent deep deterministic policy gradient(MADDPG)
mission planning
作者简介
王桢朗(1998-),男,硕士,主要研究方向为任务规划、深度强化学习、卫星应用;通信作者:周军(1982-),男,硕士,研究员,主要研究方向为卫星综合电子、卫星应用。