Year: 2023
Type: dissertation
Abstract: In-situ exploration on asteroid terrain surfaces with help of a multicopter-like spacecraft permits versatile on-site scientific observation and investigation. Relocation manoeuvres in hovering mode require autonomous and accurate navigation for spacecraft state estimation and terrain mapping. The Guidance-Navigation-Control (GNC) system in place in such a mission is confronted with specific technical and environmental challenges on the asteroid surfaces. In this dissertation, a navigation architecture with a minimum sensor setup based on a flash LiDAR-aided inertial navigation system is proposed for exploration of the asteroid terrain surfaces. Among a variety of different alternatives for the main navigation sub-functions: state estimator, point cloud-based registration and terrain mapping, three navigation variants have been investigated in detail and traded against each other. 1. Error state Kalman filter (ESKF)-based global relative navigation with iterative closest point (ICP) and extended digital elevation mapping (eDEM), 2. ESKF-based local relative navigation with scan-to-map matching (S2M)/ICP and surfel grid mapping (SGM), 3. Optimization-based local relative navigation with ICP and SGM. In a high-fidelity simulation environment for the asteroid terrain surface, the three navigation variants are tested, compared and evaluated in several test scenarios. Furthermore, factors (surface roughness, flight height, flash LiDAR inclination and flash LiDAR field of view) that affect the navigation performances are studied. To complement the navigation system, a path planning algorithm using the SGM, a rock point cloud extraction and description method are proposed and validated in this dissertation as well. This dissertation delivers ultimately insights of properties of the different navigation solutions and lessons-learned of impacts of the factors on the navigation performances. Based on the insights and lessons-learned, a key question in navigation design in context of this dissertation is answered: Which navigation configuration with which characteristics is suitable for a certain specific mission?:Abstract Kurzfassung 1. Introduction 2. State-of-the-Art and Contribution 3. Basics 4. State Estimation 5. Point Cloud-based Registration 6. Terrain Mapping 7. Local Point Cloud Descriptor 8. Performance Evaluation 9. Conclusion and Outlook Appendix
Author: Liu, Bangshang
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7015381322