Abstract
The increasing amount of space debris is posing a serious threat to the sustainable use of
space. ADR is a promising solution that involves capturing and disposing of debris using
a chaser spacecraft. The success of such missions depends on accurate relative navigation
between the chaser and the debris. In the case of an unknown and uncooperative piece of
debris, relative pose determination is not possible as no reference is available beforehand.
This thesis proposes an onboard software architecture for estimating the shape and state
of an unknown space debris, as well as for tracking it, using features detected in the
image and the associated depth.
The backend, responsible for performing the estimation, is first introduced. An ESKF
is used to estimate the chaser trajectory and attitude, as well as the target shape in an
arbitrarily fixed target frame. Next, GPs algorithm is applied to model the rotational
dynamics of the debris, enabling an accurate attitude propagation independently of its
tumbling mode. Finally, a second ESKF is instantiated using existing knowledge to track
the position and attitude of the target.
The thesis then presents the front-end, responsible for processing images and associating
depth data to retrieve and track salient features between images. The commonly
used gradient corner detectors and KLT trackers are employed to perform these operations.
The implementation of a reliable loop-closure mechanism motivates the development
of a meta-matcher consensus algorithm, leading to a drastic decrease in false
positive matching compared to commonly used feature-matching algorithms.
The entire pipeline, including the front-end and back-end, is then tested on images
and depth data generated using a software simulator and a robotic testbed specifically
adapted for this work.