【摘要】 提出了一种在轨道空间中卫星集群中进行的非合作目标追踪和逃逸博弈的方法。该方法解决了复杂的动态模型和卫星之间难以协调的问题。基于多智能体深度强化学习算法，首先构建了卫星集群博弈场景的动态模型，并通过多智能体深度确定性策略梯度算法（MADDPG）训练每个卫星的最优策略，同时考虑最小燃料消耗和最短时间，然后使用分布式执行来实现追逐和逃逸博弈。结果表明，当追逐和逃逸卫星的性能相同时，它们之间的距离保持恒定，达到纳什均衡点；当逃逸的航天器、追逐卫星的数量和性能不同时，追逐卫星可学习最优策略并成功捕获逃逸的航天器。该方法可为解决轨道空间中卫星集群的博弈问题提供一种新的方法和途径。更多还原

【Abstract】 A method for non cooperative target tracking and escape game in satellite clusters in orbital space is proposed. This method solves the problem of complex dynamic models and difficult coordination between satellites. Based on the multi-agent deep reinforcement learning algorithm, a dynamic model of the satellite cluster game scenario was first constructed, and the optimal strategy for each satellite was trained using the multi-agent deep deterministic policy gradient algorithm(MADDPG),while considering the minimum fuel consumption and shortest time. Then, distributed execution was used to achieve the chase and escape game. The results indicate that when the performance of chasing and escaping satellites is the same, the distance between them remains constant, reaching the Nash equilibrium point. When the number and performance of escaping spacecraft and chasing satellites are different, chasing satellites can learn optimal strategies and successfully capture escaping spacecraft. This method can provide a new approach and approach for solving the game problem of satellite clusters in orbital space.更多还原