A Robust Monocular Line-based Relative Navigation Approach for Spacecraft Rendezvous and Proximity Operations

Abstract

Simultaneous Localization and Mapping (SLAM) algorithms have demonstrated considerable efficacy in ground robotics applications. However, their application to space-based scenarios is constrained by many challenges, such as harsh lighting, large distances between the observer and the target, as well as on-board computational limitations. With the goal of enhancing resilience against conditions encountered in orbit, we incorporate geometric primitives into a previously developed visual SLAM pipeline for the space environment and create a sophisticated keyframe selection metric, that is able to identify the keyframes online in a smart way. These modifications render our algorithm resilient to lighting variations and to the sudden or slow motion of the spacecraft. The algorithm is tested on synthetic datasets generated specifically to capture these characteristics, as well as real-life imagery from a NASA proximity operation to the Hubble Space Telescope.