Lidar-Based Relative Pose Estimation Method

Pose estimation of non-cooperative spacecrafts confronts numerous challenges in space missions. Especially when handling weak-texture targets, traditional visual sensors are constrained by their dependence on features. When the target surface lacks distinct textures or has sparse features, it is difficult for them to provide accurate pose information. In contrast, Lidar can overcome the difficulties caused by low-texture or sparse-feature targets, thereby achieving more precise 3D pose estimation. Addressing the problem of Lidar-based 3D-3D pose estimation, this paper proposes a globally optimal pose estimation method based on the Branch-and-Bound (BnB) algorithm. Specifically, we first construct maximum consensus set-based cost functions for the rotation matrix and translation vector, respectively. Subsequently, we derive the upper and lower bounds of the rotation matrix-oriented cost function within the BnB framework. We then iteratively execute branching, upper/lower bound evaluation, and pruning operations until the algorithm converges, thereby acquiring the globally optimal solution of the rotation matrix. For the translation vector, a Huber loss function is formulated, which is further solved through an iterative approach. Finally, extensive experiments are conducted on synthetic datasets to validate the effectiveness and superiority of the proposed algorithm.