Design and testing of a vision based navigation system for a spacecraft formation flying simulator
Year: 2018
Type: dissertation
Abstract: In the last decades, Spacecraft Formation Flight (SFF) and proximity operations have aroused a great deal of interest within the international space community for the wide range of advantages generated with respect to traditional mission architectures. The idea of distributing the capabilities of space systems over multiple satellites, in general, results in lower development costs, redundancy to failure, flexibility and improved measurement performances of the payloads. Moreover, in sight of the extension of the operational lifetimes of orbiting spacecraft, performing on orbit inspections, maintenance and repair provides significant advantages from an economical point of view. Nonetheless, all these advantages comes at a cost as the execution of coordinated maneuvers between two or more spacecraft impose requirements not only on the absolute state, but also on the relative dynamic state. This reason pushes researchers and institutions to advance, from a technical and technological point of view, the development of accurate and robust Guidance, Navigation and Control (GN&C) strategies to enable the autonomous execution of relative navigation tasks on orbit. Prior to be accepted for the flight, each strategy has to be validated in a relevant and representative environment: one popular and effective way to reach this objective is means of ground-based testbeds. The current work focuses on the development of vision based estimation strategies for relative navigation tasks in the context of Spacecraft Formation Flight and proximity operations. In order to validate the developed GN&C strategies in a relevant environment, part of my activities are focused on the development and realization of the cooperating spacecraft testbed for autonomous proximity operations experiments, called SPARTANS, whose design and realization started in 2010 at the Center of Studies and Activities for Space (CISAS) "G. Colombo" of the University of Padova. SPARTANS is a ground-based hardware simulator that features robotic units that allow to reproduce the relative motion of two ore more spacecraft flying in formation or in close proximity, allowing therefore to validate advanced relative GN&C strategies developed within this context. The first activity of my Ph.D. regards the characterization of a monocular vision system for relative pose estimation in the context of cooperative spacecraft. An original custom-made set of fiducial markers is proposed along with the required image analysis procedure to extract relevant points from the acquired images. The estimation of the relative pose is obtained by solving the perspective-from-three points problem followed by a subsequent pose refinement procedure. The metrological performances of the proposed approach are assessed following an experimental approach: known displacements are imposed to a satellite mock-up and, by comparing the estimated poses with the imposed ones, the measurement error is characterized. Finally an uncertainty analysis, validated by a series of tests, was performed yielding an average accuracy of about 1 cm and 1 deg on the relative position and attitude, respectively. This activity is followed by the implementation of new features for the SPARTANS testbed. At the beginning of my Ph.D., the testbed featured only one spacecraft module capable of performing 3 degrees of freedom attitude maneuvers exploiting the measurements coming from incremental quadrature encoders and actuating control actions thanks to a cold gas thrusters system. A small contribution is initially given in the development of the Translational Module (TM), that enables low friction planar displacements to the robotic platforms, and in the design and development of the laboratory test table that provides the testbed with a planar surface on top of which the spacecraft simulators can translate. Subsequently, I was involved in the realization, from a practical point of view, of the laboratory setup: in particular, the second main activity I have been working on regards the development, the implementation and the characterization of a global navigation system that allows to estimate the position and the orientation of each spacecraft module within the laboratory operative volume. The first step of this activity consists in the development of a cost-effective prototype based on a monocular system and a set of three Optical Flow Sensors to obtain planar pose estimations of the TM. The estimation scheme relies on two sets of measurements that are conveniently merged by means of an Extended Kalman Filter. The second development phase consists in the extension of the proposed estimation approach by adapting it to the laboratory characteristics: a Motion Capture system consisting of six infrared cameras is employed to this purpose. A pose estimation strategy is proposed in this context and a proper calibration procedure is developed. The results of the characterization of the Motion Capture system shows sub-degree and sub-millimeter accuracy respectively on the estimated orientations and positions, allowing to obtain reference measurements to assess the performances of the GN&C algorithms tested with the spacecraft simulator. The last part of my activity is focused on the development of a relative navigation strategy based on an Extended Kalman Filter that exploits only visual data provided by a stereoscopic vision system. The proposed estimation scheme relies on the a priori knowledge of the geometry of the observed object (target) and it is implemented in C++ in sight of a foreseen implementation on embedded systems. The filter capabilities are assessed by means of numerical simulations in two different scenarios. In the first scenario, the relative dynamic state of two spacecraft orbiting in a leader-follower formation at short range is simulated numerically by means of a software orbital propagator. In the second scenario, the operative conditions of the laboratory setup are reproduced by means of a 3D simulated environment. Finally, the proposed navigation strategy is validated by means of laboratory testing exploiting the robotic spacecraft units. The proposed navigation algorithm proves to be capable of estimating the relative dynamic state of an observed target with centimeter-level and degree-level accuracy. In its current configuration, the SPARTANS testbed features a first complete spacecraft unit, a 3 x 2 m flat test table, a global navigation system consisting of 6 infrared cameras and an external control station. Once the construction of the second spacecraft unit will be completed, the testbed will allow to perform coordinated maneuvers between two (or more) spacecraft units for the study and validation of strategies related to SFF, automated rendez-vous and docking and, in general, proximity maneuvers.
Author: Mattia Mazzucato
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2908649888