Answers are grounded in this paper and cite their sources.
Horizon-based Optical Navigation (OPNAV) is one of the most important Celestial Navigation Systems (CNS) for deep space exploration missions. Compared with the traditional navigation method, which relies on ground-based tracking and inter-spacecraft measurements, the OPNAV realizes autonomous onboard navigation using centroid and apparent-diameter measurements and an ellipse-fitting algorithm. However, the accuracy of the OPNAV relies on the initial value, including the initial state velocity and attitude. and the result becomes divergent when a large initial error is present. Additionally, all cameras cannot always capture target figures in Halo orbit at the same time, which is a key issue for OPNAV. To address these issues, this paper presents an autonomous navigation system that relies on a star tracker and two horizon sensors, including an Earth sensor and a Moon horizon sensor. The star tracker can provide high-accuracy attitude information to improve the initial attitude and state from two horizon sensors based on the Quaternion-Estimator Algorithm (QUEST). The two-group fusion results from the QUEST algorithm will be combined using the Unscented Kalman Filter(UKF) to enhance navigation accuracy. The proposed algorithm uses three Liquid Neural Networks(LNNs) to maintain navigation accuracy if one of the sensors loses target lock. This system can provide high autonomous navigation information without the ground tracking at the Earth-Moon L 2 point.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: