ESA GNC Conference Papers Repository

For autonomous on-orbit servicing, knowledge about the target spacecraft's 3D shape during the proximity phase is essential for collision-free interaction and servicing. In case the target has an unknown shape (e.g. space debris, damaged satellite) the 3D structure has to be estimated. Camera based solutions offer advantages in costs, weight and power consumption, but require proper algorithms for evaluation. Usual 3D shape reconstruction algorithms are based on point clouds with considerably high point densities. But for real-time and on-board estimation only sparse target information is manageable. We present a new approach estimating the 3D shape of an uncooperative and unknown target spacecraft (S/C) based on a sparse point cloud obtained from camera image data. First a S/C-Rendezvous-SLAM (Simultaneous Localization and Mapping) algorithm estimates the sparse 3D point cloud of the target's surface and the relative kinematic states simultaneously. The resulting noisy point cloud is used to reconstruct the unknown target 3D structure to allow collision avoidance and also a geometrically characterization of the target. For validation and verification of the approach a spacecraft rendezvous simulator was developed and is presented in this paper.