City Research Online

Abstract

The growing necessity for autonomous space operations has intensified due to the proliferation of on-orbit servicing missions and the critical need to mitigate space debris accumulation, highlighting the essential role of precise and reliable autonomous docking systems. In response to these challenges, this paper presents and validates a novel hybrid methodology for autonomous spacecraft docking that integrates Convolutional Neural Networks (CNNs) with Perspective-n-Point (PnP) algorithms for monocular pose estimation. The proposed hybrid framework synergistically combines CNN-based keypoint detection with PnP geometric reconstruction and RANSAC-based outlier rejection to achieve robust and accurate pose estimation under diverse operational conditions, including variable illumination, viewing geometries, and approach trajectories. A comprehensive evaluation of CNN backbone architectures was conducted using both synthetic and real-world datasets to optimize performance characteristics, encompassing ResNet50, MobileNet, EfficientNet, and HRNet architectures. Experimental validation was performed in a controlled facility utilizing robotic hardware and specialized illumination systems designed to replicate space environmental conditions. The system demonstrated exceptional performance, maintaining translational errors below 0.30% and rotational errors below 1.14º during simulated docking scenarios. Comparative analysis with other direct pose estimation methodologies confirms that the proposed hybrid approach achieves superior translational accuracy while preserving high rotational precision, establishing its viability for autonomous spacecraft operations.

Publication Type:	Article
Additional Information:	© 2025 The Authors. Published by Elsevier Ltd on behalf of IAA. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Publisher Keywords:	Deep learning, Hybrid pose estimation, Computer vision, Autonomous docking, Space robotics, Guidance, navigation and control
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QB Astronomy T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TL Motor vehicles. Aeronautics. Astronautics
Departments:	School of Science & Technology School of Science & Technology > Department of Engineering
SWORD Depositor:	Symplectic Administrator

Export

Downloads