6D Pose Estimation and Calibration for Autonomous On-Orbit Servicing using the OOS-SIM

Abstract

Reliable six-degree-of-freedom pose estimation is a key requirement for au- tonomous on-orbit servicing (OOS), in which a robotic spacecraft performs tasks such as inspection, maintenance, and life extension of other satellites in orbit. However, generating trustworthy ground truth for algorithm develop- ment remains challenging. This thesis investigates calibration, sensor evalu- ation, and sensor fusion for the OOS-SIM testbed at the German Aerospace Center, which combines industrial robots, a VICON motion capture system, and forward kinematics (FK) to emulate servicer–client satellite interactions. The thesis first formulates the system calibration as an AX = YB hand–eye problem and employs a linear solver to align VICON and FK measurements in a common world frame, achieving sub-millimetre and sub-tenth-degree cal- ibration residuals. A camera-based ground truth concept was then explored using a calibra- tion board observed by an end-effector stereo camera. In practice, exposure limitations, motion blur, and likely changes in camera intrinsics prevented the camera system from providing a reliable reference, and its measurements could not be trusted to outperform VICON or FK. Finally, an Invariant Ex- tended Kalman Filter was implemented to fuse VICON and FK poses on SE(3). The filter exhibited stable and interpretable behaviour, correctly re- flecting the chosen sensor covariances, but its benefits were constrained by the lack of a motion model and the absence of dependable ground truth under challenging conditions. Overall, the thesis highlights the critical role of data integrity, calibration design, and experimental setup in establishing meaning- ful pose ground truth for space robotics research.