Factor-Graph-Based Camera/Lidar Fusion for Defunct Spacecraft Attitude Estimation

To address the estimation accuracy degradation due to insufficient computational performance of satellite platforms that cannot meet the high-frequency sampling requirements of visual sensors, a camera/LIDAR fused relative navigation method is proposed within a graph optimization framework by exploiting the conservation of momentum. Firstly, measurement data preprocessing is conducted, where an elliptical arc recognition algorithm is used to detect circular visual features on the surface of the defunct spacecraft, obtaining camera attitude observation. Simultaneously, the Iterative Closest Point (ICP) algorithm is employed to acquire LIDAR attitude measurement data. Secondly, pseudo-measurement factors are designed for both the camera and LIDAR, and a variational integration factor is designed for the attitude increment between adjacent moments of the defunct target. Finally, a sliding-window-based factor graph optimization is performed, maintaining high estimation accuracy with lowfrequency measurements, to achieve the Maximum A Posteriori (MAP) estimation of the relative attitude of the defunct noncooperative spacecraft.