Reconstruction of the shape of a tumbling target from a chaser in close orbit

Operations involving two or more spacecraft, including approach, rendezvous and servicing, are not always cooperative. The lack of cooperation means a limited set of information initially available to the approaching spacecraft. Still, the determination by the chaser of the relative kinematic state of the target spacecraft stands as a required step to continue the approach. Completing this first fundamental information about relative motion, also the reconstruction of the target's shape can be considered an important part of the rendezvous and a pre-requisite for safe docking. In fact, shape reconstruction enables the chaser to understand target's configuration, to assess its integrity and eventually to compare it with already known spacecraft's models. The reconstruction needs to be accurate even when starting from the limited number of images captured only from the points of view attained during relative motion, and is indeed quite a challenging task. Due to the extremely wide range of possible relative poses and light conditions, an extensive test campaign based on numerical simulations is required to validate possible algorithms to carry on this operational phase. Only after simulations are successfully passed, would it be possible to move towards experiments in a ground-based testbed and finally to in-flight qualification. This paper details the experience gained with the simulation phase at the Guidance and Navigation Lab of Sapienza University of Rome. A software suite is developed in order to simulate in orbit acquisition of the target image, managing its 3D CAD model according to the relative dynamics and lighting conditions. Several spacecraft configurations, different in terms of shapes and relative pose, are assumed for the target, and the relevant images are captured by the chaser during its relative motion around it. An effective process for the identification and match of the features, capable to manage their appearance and disappearance during the sequence of images has been implemented. Following these steps, it is possible to gain an educated guess of the shape of the target. Advanced filtering techniques are applied, significantly contributing to the final result of the process. The post-facto comparison between the actual CAD model and the estimated target's shape shows appealing success rates in the recognition task for the proposed technique