面向空间防御的多航天器协同博弈方法OA

To address the cooperative observation and interception problem in multi-spacecraft space defense scenarios,this paper proposes a hierarchical intelligent cooperative architecture that integrates game theory and reinforcement learning.The architecture decouples observation task allocation and maneuver control:the upper-level observation game layer employs a Stackelberg game algorithm,which synthesizes solar angle constraints,observation distance,threat assessment,and information entropy for time-sensitive cooperative observation assignment.The lower-level control execution layer adopts the Multi-Agent Proximal Policy Optimization(MAPPO)algorithm,following a centralized training with decentralized execution paradigm,to realize autonomous interception control that complies with orbital dynamics and thrust constraints.To simulate the uncertain behaviors of non-cooperative spacecraft,a stochastic policy model is proposed based on the Clohessy-Wiltshire equations,by combining a deterministic baseline policy with Gaussian perturbations.Enabled by information sharing and fusion mechanisms,the system achieves a"one-point observation,global awareness"cooperative perception mode,significantly enhancing situational awareness.Simulation results demonstrate that the proposed hierarchical architecture effectively reduces computational complexity while improving system scalability and decision-making efficiency,and ensures high interception success rates with low fuel consumption.Thus,it provides a novel approach for multi-spacecraft cooperative missions in space confrontation scenarios.