Abstract: To address the challenges of unknown target intent and strategy selection under incomplete-information orbital games, an intent-inference-based maneuver-decision framework is proposed. Exploiting the impulsive characteristics of spacecraft, a model-predictive-control scheme is first devised within a finite-time horizon to rapidly generate guardian strategies for a single prespecified intent. Subsequently, an integrated guardian maneuver optimization method is developed that fuses probabilistic intent-inference outputs, thereby extending applicability to multi-intent scenarios. The optimization objective is dynamically re-weighted according to the inferred intent distribution, affording enhanced adaptability to uncertain operational environments. Simulation results across diverse space-guardian missions confirm the superior performance of the proposed approach.

Key words: Non-cooperative target, Intent Inference, Deep Learning, Spacecraft Guarding, Orbital Game