Dual-Layer Factor-Graph Optimization for Delayed Star-Tracker/IMU Fusion in Highly Dynamic Spacecraft Attitude Estimation

Accurate attitude estimation for highly dynamic spacecraft relies on robust fusion of star-tracker and inertial measurements. However, asynchronous sensing, motion blur in star images, and delayed star-tracker outputs can significantly degrade estimation accuracy and temporal consistency. To address these challenges, this paper proposes a dual-layer factor graph optimization framework for asynchronous star-tracker/IMU fusion under highly dynamic conditions. At the lower layer, high-rate IMU measurements are combined with motion-blurred star streak observations to construct a local factor graph over the exposure interval. The proposed local fusion process reconstructs discrete star-trail points, estimates angular velocity, and selects IMU-aligned representative observations for temporally consistent association of blurred star measurements. At the upper layer, delayed attitude constraints, propagated star-vector information, and inertial rotational constraints are jointly incorporated to refine the attitude trajectory. Simulation and semi-physical experimental results demonstrate that the proposed framework achieves higher estimation accuracy, stronger robustness, and better tolerance to delayed or intermittent star-tracker observations than the comparison methods, while maintaining practical computational efficiency for near-real-time onboard implementation.