Spacecraft Perception
Credit: European Space Agency, License: ESA Standard License
Introduction
Sustainable space utilisation demands space missions that are reusable, extendable and amenable to safe end-of-mission disposal. On-orbit servicing (OOS) encompasses a range of important capabilities that support sustainable space utilisation. Major applications enabled by OOS include satellite maintenance (e.g., refuelling, repairing, upgrading), constructing larger space installations (e.g., space stations) by aggregating smaller modules in space, and removing space debris and defunct satellites from orbit.
Spacecraft RPO and autonomy
Technologies contributing to OOS have been steadily progressing over many decades. Rendezvous and proximity operations (RPO) (e.g., rendezvous, docking, robotic interactions) are critical in enabling multiple spacecraft to come within close distances and interact with each other to perform OOS. However, to be viable, commercial OOS must be reliable and scalable, i.e., running multiple concurrent missions per year. This necessitates the usage of spacecraft with a high level of autonomy, such that delicate RPOs can be conducted safely with as little human oversight and intervention and as possible.
Spacecraft perception
Our lab develops cutting edge spacecraft perception algorithms that play a critical role in autonomy for OOS. Important capabilities enabled by spacecraft perception include relative pose estimation, 3D reconstruction, object recognition (e.g., identifying satellite of interest), and semantic understanding (e.g., classifying spacecraft types and parts).
Neuromophic vision for RPO
Our lab pioneered the usage of neuromorphic sensing and vision for spacecraft perception to conduct RPOs in OOS applications. Since neuromorphic sensors have much higher dynamic ranges, they would be much less affected by the harsh lighting conditions in space (e.g., significant brightness or dimness).
Relevant publications
Abdul Mohsi Jawaid, Rajat Talak, Luca Carlone, Tat-Jun Chin. Test-Time Certifiable Self-Supervision to Bridge the Sim2Real Gap in Event-Based Satellite Pose Estimation. IROS 2024.
Ethan Elms, Yasir Latif, Tae Ha Park, Tat-Jun Chin. Event-based Structure-from-Orbit. CVPR 2024.