Recently, the paper "An Algorithm for Affordable Vision-Based GNSS-Denied Strapdown Celestial Navigation" came across Hacker News (discussion link),
This post is focused on research starting from the paper but focusing on nearby topics. A separate post will review the paper in more detail (An Algorithm for Affordable Vision-Based GNSS-Denied Strapdown Celestial Navigation, A Review)
Citation: Teague, S.; Chahl, J. An Algorithm for Affordable Vision-Based GNSS-Denied Strapdown Celestial Navigation. Drones 2024, 8, 11.
Paper References¶
I found the paper's references to be useful in addition to finding the paper interesting itself, so I've focused this post on interesting citations and adjacent topics as opposed to the paper itself.
The Math¶
Van Allen, J.A. Basic principles of celestial navigation. Am. J. Phys. 2004, 72, 1418–1424. Link
This paper was cited for the math for projecting star observations onto the terrestrial sphere
Other Celestial Methods¶
Wang, J.; Chun, J. Attitude determination using a single star sensor and a star density table. J. Guid. Control Dyn. 2006, 29, 1329–1338.
This paper was cited as an example of a space application that uses celestial positioning for attitude reference
Simulation¶
Teague, S.; Chahl, J. Imagery synthesis for drone celestial navigation simulation. Drones 2022, 6, 207. DOI Review
This paper was used to simulate measurements in the parent paper, including testing against motion blur effects and testing the effect of wind conditions on accuracy.
I have a few stacked interests in this topic. First, I'd like to create simulated imagery of celestial bodies and re-implement the paper. Second, I'd like to use the imagery to generate a synthetic map of the celestial sphere and connect that to compressive sensing topics to create a compact representation. From there I'd like to extend to position estimation and performing that in a simulated environment. N.B. The imagery synthesis also covers converting a star's database information into the theoretical observed position
Star Databases¶
Wei, X.; Zhang, G.; Jiang, J. Star identification algorithm based on log-polar transform. J. Aerosp. Comput. Inf. Commun. 2009, 6, 483–490.
From the paper: "During the instantiation of the star tracker, a lost-in-space log-polar star identification algorithm is used to determine the IDs of each star in the frame". This initialization of the star tracker and generic star identification seems quite useful to run in a parallel loop as a kidnapped robot recovery process. As an extension of this parallel process, a more holistic approach to this could be to manage the estimation as a multiple-hypothesis filter, where the lost-in-space algorithm generates a new hypothesis at a regular cadence and the top N most likely hypothesis are maintained for ongoing estimation.
In addition to what's cited by the original paper, I also found a recommendation for the HYG Database (also on Codeberg). From Codeberg
HYG combines every identifiable star in the HIPPARCOS, Yale Bright Star, and Gliese (nearby star) catalogs into a combined dataset of the stars' currently best-known positions, brightnesses, spectral types, and various additional catalog IDs such as traditional names and Bayer designations.
Tools for thought related papers¶
This content was surfaced by integrating with my own tools that search arXiv.org to help find interesting and related content.
History of Celestial Navigation¶
The beginning of celestial navigation https://arxiv.org/abs/2209.02371v1
Celestial Navigation System Design¶
Conceptual Design on the Field of View of Celestial Navigation Systems for Maritime Autonomous Surface Ships https://arxiv.org/abs/2408.15765v1
Orbit Estimation Using a Horizon Detector in the Presence of Uncertain Celestial Body Rotation and Geometry https://arxiv.org/abs/1804.04401v2
Compressive Sensing for Efficient Representation¶
Compressive Sensing with Local Geometric Features https://arxiv.org/abs/1208.2447v1
Taking from the abstract:
We propose a framework for compressive sensing of images with local distinguishable objects, such as stars, and apply it to solve a problem in celestial navigation. Specifically, let x be an N-pixel real-valued image, consisting of a small number of local distinguishable objects plus noise. Our goal is to design an m-by-N measurement matrix A with m << N, such that we can recover an approximation to x from the measurements Ax.
This seems like an immediately useful application for finding an efficient representation of different views of the sky. Without the tools for thought script, I was only peripherally aware of compressive sensing and would not have thought that it related to my interest in celestial navigation.
CELESTIAL: Classification Enabled via Labelless Embeddings with Self-supervised Telescope Image Analysis Learning https://arxiv.org/abs/2201.08001v1
This paper covers using extensive (petabytes) of unlabelled data to learn a [compressed?] representation of the image class, which feels like it could be useful for learning a compressed representation of star observations.
Space Vision Applications¶
AstroVision: Towards Autonomous Feature Detection and Description for Missions to Small Bodies Using Deep Learning https://arxiv.org/abs/2208.02053v1
This paper seems interesting if a bit off-topic for Earth-based navigation; however, the visual navigation feels like it would overlap with my other visual-odometry research interests
Additional Special Issue Topics¶
Paper by the MDPI journal Drones, specifically a special issue on Drones Navigation and Orientation
... Written with tools for thought to help connect to new ideas I wouldn't have found otherwise