Vision-based autonomous characterization and navigation for asteroids

Abstract

The research effort aims at designing a solely vision-based process capable of esti- mating the rotational dynamics and prelimi- nary shape of an asteroid, with the prospect of being embedded on a probe, which is a completely novel approach. A robotics- inspired method of Extended Kalman Fil- ter Simultaneous Localization and Mapping (EKF-SLAM) is proposed as a suitable first step towards an ultimately autonomous char- acterization scheme. A 2D feature extraction technique is applied to a monocular camera image stream of the asteroid surface in the visible spectrum, producing a bearing-only measurement input for the estimation engine. Depth information is inferred through triangulation of time-delayed bearings. Applicability of the method has been verifieed through promising results obtained with a simulated environment, as well as a test bench exploiting a camera to produce measurements of an actual moving target.