Spacecraft Pose Estimation: Robust 2-D and 3-D Structural Losses and Unsupervised Domain Adaptation by Intermodel Consensus
Juan Ignacio Bravo Pérez-Villar, Álvaro García‐Martín, Jesús Bescós, Marcos Escudero-Viñolo
Abstract
The accurate estimation of spacecraft pose is crucial for missions involving the navigation of two spacecraft in close proximity. Supervised algorithms are currently the state-of-the-art approach for spacecraft pose estimation. However the absence of training data acquired in operational scenarios poses a challenge for the supervised algorithms. To address this issue, computer-aided simulators have been introduced to solve the issue of data availability but introducing a large gap between the training domain and test domain. We here describe an algorithm for unsupervised domain adaptation with robust pseudo-labelling by model consensus. Moreover, the proposed method incorporates 3D structure into the spacecraft pose estimation pipeline to provide robustness against high illumination shifts between domains. Our solution has ranked second in the two categories of the 2021 Pose Estimation Challenge (SPEC2021) organised by the European Space Agency and the Stanford University, achieving the lowest average error over these two categories. Our solution is available at: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/JotaBravo/spacecraft-uda</uri>