Napolano, Giuseppe (2025) Advanced GNC Solutions for Autonomous Spacecraft Proximity Operations. [Tesi di dottorato]

	Documento PDF Napolano_Giuseppe_38_COMPLETO.pdf Visibile a [TBR] Amministratori dell'archivio Download (17MB) | Richiedi una copia
	Documento PDF Napolano_Giuseppe_38_PARZIALE.pdf Visibile a [TBR] Amministratori dell'archivio Download (20MB) | Richiedi una copia

Abstract

This Ph.D. thesis addresses the challenge of enabling autonomous Guidance, Navigation and Control functionality for satellites performing close-proximity operations with resident space objects in the context of on-orbit servicing or active debris removal missions. In order to guarantee accurate relative navigation capability during inspection and rendezvous phases, the thesis investigates the use of monocular sensors on the autonomous spacecraft (chaser) to determine relative position and attitude with respect to the space object (target), as well as to estimate its unknown angular motion. Firstly, a multi-sensor architecture is conceived to enable relative navigation towards a cooperative and stabilized target equipped with circular fiducial markers. The architecture, which can be embedded in a relative navigation module composed of 2 CubeSat units, relies on a monocular camera and a laser range finder, whose measurements are fed into a Multiplicative Extended Kalman Filter to retrieve the complete relative navigation state. Successively, an image processing algorithm is developed to extend these functionalities to uncontrolled tumbling targets; to this aim, monocular images are processed to detect and recognize polygonal markers which distinguish different approach faces. The thesis also presents techniques for relative navigation with respect to uncooperative targets. A multi-sensor relative navigation architecture, combining data from EO sensors for inspection and approach operations towards both stabilized and tumbling targets is presented. Monocular-based image processing foresees two operative modes, based on the estimation of the target’s line of sight and on the extraction of natural features on its surface for relative position and attitude (i.e., pose) determination, depending on the relative distance between the satellites. A second technique for uncooperative spacecraft relative navigation is also described, relying on convolutional neural networks for pose estimation and target angular velocity estimation; the method foresees the use of networks for target detection and for the identification of its natural features, and consistency checks are integrated in the pose estimation pipeline to enhance the robustness of the pipeline. Successively, relative attitude measurements are employed to determine the angular velocity of the target, thus making the method suitable for uncontrolled tumbling spacecraft. The proposed methodologies are tested in numerical simulation environments, reproducing sensor performance and relative motion in the orbital environment. Experimental tests are also conducted for the evaluation of the performance and of the computational cost of the cooperative relative navigation pipelines, while synthetic image datasets are generated to validate the uncooperative relative navigation algorithms based on convolutional neural networks. Considering the importance of accurate and continuous relative navigation functionality for the safety of proximity operations, a motion planning pipeline is developed to improve the performance of visual-based relative navigation during rendezvous and capture of space targets. The method relies on the definition of a set of metrics to characterize the expected accuracy of monocular-based relative navigation, and on their integration within an optimization algorithm for trajectory generation. The algorithm is tested on reference approach scenarios to a target satellite, and a procedure for trajectories’ simulation in experimental facilities, using robotic manipulators, is also implemented. Finally, in order to provide an effective application of the control commands generated by spacecraft control functions, the thesis proposes an algorithm for their distribution to a set of reaction control thrusters. The method accounts for several operative limitations on the use of the actuators, including constraints on the minimum firing time, on the maximum number of simultaneously active thrusters and on the duration of prolonged firings. The method is integrated within a numerical environment for the simulation of a Guidance, Navigation and Control subsystem, and it is validated through parametric analyses and comparison with a non-optimized algorithm.

Downloads

Actions (login required)

Modifica documento