Vision Based Navigation for Autonomous Cooperative Docking of CubeSats

A realistic rendezvous and docking navigation solution applicable to CubeSats is investigated. The scalability analysis of the ESA Autonomous Transfer Vehicle Guidance, Navigation & Control (GNC) performances and the Russian docking system, shows that the docking of two CubeSats would require a lateral control performance of the order of 1 cm. Line of sight constraints and multipath effects affecting Global Navigation Satellite System (GNSS) measurements in close proximity prevent the use of this sensor for the final approach. This consideration and the high control accuracy requirement led to the use of vision sensors for the final 10 m of the rendezvous and docking sequence. A single monocular camera on the chaser satellite and various sets of Light-Emitting Diodes (LEDs) on the target vehicle ensure the observability of the system throughout the approach trajectory. The simple and novel formulation of the measurement equations allows differentiating unambiguously rotations from translations between the target and chaser docking port and allows a navigation performance better than 1 mm at docking. Furthermore, the non-linear measurement equations can be solved in order to provide an analytic navigation solution. This solution can be used to monitor the navigation filter solution and ensure its stability, adding an extra layer of robustness for autonomous rendezvous and docking. The navigation filter initialization is addressed in detail. The proposed method is able to differentiate LEDs signals from Sun reflections as demonstrated by experimental data. The navigation filter uses a comprehensive linearised coupled rotation/translation dynamics, describing the chaser to target docking port motion. The handover, between GNSS and vision sensor measurements, is assessed. The performances of the navigation function along the approach trajectory is discussed.