Lidar Navigation for Non-Cooperative Space Rendezvous
Kaydedildi:
| Yayımlandı: | PQDT - Global (2025) |
|---|---|
| Yazar: | |
| Baskı/Yayın Bilgisi: |
ProQuest Dissertations & Theses
|
| Konular: | |
| Online Erişim: | Citation/Abstract Full Text - PDF Full text outside of ProQuest |
| Etiketler: |
Etiket eklenmemiş, İlk siz ekleyin!
|
| Özet: | To ensure a sustainable and ambitious future use of space, it is essential to maintain the growing satellite infrastructure, and reduce the number of space debris. This is achieved by in-orbit refueling, repair, and active debris removal missions. At the core of these missions is the capability to autonomously rendezvous with a non-cooperative spacecraft. The close range phase of a rendezvous, when the inter-satellite distance is below tens of meters, is the most critical. Precise relative navigation is required to perform a safe approach and potential robotic activities, and is enabled by electro-optical sensors embarked onboard the spacecraft. Lidar sensors oer particular advantages for this task, since they provide accurate 3D point cloud measurements of the target satellite, independently of the external lightning conditions. Relative navigation then consists in matching the lidar point clouds collected onboard with a pre-defined model of the target, to retrieve a relative pose estimate.This dissertation presents the design and experimental validation of a lidar-based navigation system for non-cooperative space rendezvous. There exists precise methods for point cloud tracking, however, a major constraint is to perform pose estimation onboard and in real-time. Since space-qualified onboard computers have limited processing power, this imposes strong runtime requirements. Next to point cloud tracking, finding an initial pose estimate in real-time is also particularly challenging, since local matching method are not applicable. Many spacecraft also present a symmetrical shape, making an unambiguous pose estimation impossible. Further, spacecraft are often composed of reflective materials, which lower the quality of the point cloud data. A navigation system should be able to cope with this noise, and to handle the fast relative dynamics of an uncontrolled satellite or space debris, which might tumble with several degrees per second.Focusing on these core problems, the dissertation is organized in three blocks: The first is the development of a precise and onboard capable point cloud tracking method. A variant of the Normal Distribution Transform is developed, consisting in the application of a Gaussian smoothing to the probabilistic map representation, alongside with a relaxed formulation of the optimization problem. It is shown how this method leads to increased robustness, precision and eciency compared to the state-of-the-art. The second block is the combination of this tracker with a filter to form a robust navigation system. A new formulation of a relative navigation filter is introduced, based on expanding the Invariant Extended Kalman Filter formalism. It is used in interplay with the tracking method to provide accurate pose estimates, and to compensate for motion blur when dealing with rapidly tumbling space objects. The third segment is the design of a framework for lidar-based pose initialization. The approach relies on training a neural network with synthetic point cloud data generated by a custom rendezvous lidar simulator, and optimizing the network to achieve real-time capability. An attitude classification logic is further introduced to handle the specificities of symmetrical spacecraft.The findings of this research are supported by experiments conducted at the European Proximity Operations Simulator. The navigation system is precise and reliable when tested on real lidar data, under various scenarios and conditions. Runtime evaluations also demonstrate its real-time capability on onboard representative hardware. The lidar-based navigation system is flight-ready, laying the groundwork for the widespread use of lidar sensors in upcoming rendezvous missions. |
|---|---|
| ISBN: | 9798270212469 |
| Kaynak: | ProQuest Dissertations & Theses Global |