Spacecraft Pose Estimation

Spacecraft pose estimation, determining a spacecraft's 3D position and orientation from images, is crucial for autonomous rendezvous and proximity operations. Current research heavily focuses on mitigating the "domain gap" between synthetic training data and real-world imagery using techniques like test-time adaptation, unsupervised domain adaptation, and multi-task learning within convolutional neural networks (CNNs) and vision transformers. These advancements, coupled with efforts to optimize model efficiency for deployment on resource-constrained onboard computers, are paving the way for more reliable and autonomous space missions.