Vision-based Distributed Pose Estimation using a Spacecraft Constellation

In this paper, we present a simulation framework and consensus algorithms to track the pose (position and attitude) of an uncooperative/defunct spacecraft (Target spacecraft) using a fleet of small spacecraft (Ego spacecraft) in Low Earth Orbit (LEO). Since it is not possible to put traditional active sensors (like differential GPS, inter-satellite communication) on the uncooperative/ defunct Target spacecraft, the fleet of Ego spacecraft have to rely on cameras and inter-spacecraft communication among themselves to estimate the relative pose of the Target spacecraft. We present an end-to-end simulation pipeline for simulating high-fidelity orbital mechanics of multiple spacecraft in Earth orbit and generating photo-realistic images of the Target spacecraft using cameras onboard each Ego spacecraft. We then present a class of consensus algorithms (both in Eucledian and Riemannian manifolds) that help combine the individual pose estimates of the Target spacecraft from all the Ego spacecraft, under mild assumptions on the communication network topology. We provide numerical simulation results to prove the effectiveness of this approach.