Spacecraft rendezvous and proximity operations depend on accurate pose estimation, but conventional cameras struggle with extreme lighting. Event cameras, which respond to changes in brightness, have emerged as a robust alternative. However, they produce few events when relative motion is slow, making estimation difficult.
A new arXiv preprint proposes a recurrent vision transformer designed for this setting. By maintaining temporal memory across frames, the model can infer pose even when event streams are sparse. The work specifically targets slow relative motions, a regime where event cameras are typically at a disadvantage.
Because the abstract is truncated, the article does not provide quantitative results or implementation details. The authors position the approach as a step toward reliable vision-based navigation for RPO under challenging illumination.