Visual sensor selection for satellite swarm cooperative localization

摘要

Compact satellite swarms enable future space mission concepts including cellular architecture, disaggregation,orbital construction, and cooperative inspection. Such swarms can cooperatively determine position and orien-tation (known as pose) of their constituent satellites by collecting and sharing relative pose measurements. Thispaper presents sensor selection strategies which reduce the number of computationally expensive vision-basedpose measurements which must be collected while still maintaining accurate localization. The first method ac-counts for the impact of relative viewing geometry and solar angle on relative pose measurements. The secondmethod uses uniformly sampled random spanning trees. The third method randomly selects pose measurements.Selected relative pose measurements are then fused in a Multiplicative Extended Kalman Filter for decentralizedcooperative localization. Localization results using simulated measurement data and rendered stereo images arepresented.