Hierarchical Intelligent Optimization Decision Method for Multi-Spacecraft Pursuit-Evasion Orbital Game

The escalation of space confrontations has elevated orbital pursuit-evasion games to a prominent research area, with multi-spacecraft scenarios presenting critical challenges such as high-dimensional state modeling and solution complexity. This paper proposes a hierarchical decision-making framework based on task decoupling to overcome the bottlenecks of traditional differential games and multi-agent reinforcement learning. A global task allocation layer employs an enhanced genetic algorithm to optimize pursuit-evader matching, while a local dynamic control layer applies a Deep Deterministic Policy Gradient (DDPG) model to execute pursuit strategies. A closed-loop decision flow between the two layers is established, where global allocation delineates operational targets for local control, and local strategy feedback evaluates success rates to guide global optimization, thereby achieving stable decision-making under complex game scenarios. Simulation results in a multi-spacecraft scenario demonstrate a 97% pursuit success rate, validating the hierarchical structure's effectiveness in reducing dimensionality and enhancing performance.