Development of Cooperative Vision-Based Navigation Techniques using Neural Networks

Author

Jill Davis
Henry J. Pernicka, Missouri University of Science and TechnologyFollow

Abstract

The small satellite revolution has opened the door to a new era in satellite development. As spacecraft hardware continues to advance, traditional monolithic satellites can be replaced with cooperative swarms of smaller spacecraft. To enable an effective autonomous mission architecture, swarms of spacecraft will need mechanisms to perform relative navigation for each member. This research presents a solution for the spacecraft identification problem using a neural network architecture to classify space objects and then applies traditional image processing techniques to perform pose estimation using monocular camera images. The trained neural network classifies multiple CubeSats detected in an image into a predefined number of classes using object pixel count and centroid information. The pose estimator then uses the classification to load a predefined wireframe object model and employs an iterative least squares approach to solve the inverse perspective-n-point problem. Preliminary results using a data set with 51,840 points show neural network classification accuracy of ~97% with network training times of less than ten minutes. The position error for spacecraft with a relative range of ~50 meters is typically less than one meter in the X and Y directions and around 3 meters in the Z direction.

Recommended Citation

J. Davis and H. J. Pernicka, "Development of Cooperative Vision-Based Navigation Techniques using Neural Networks," Proceedings of the IEEE Aerospace Conference Proceedings (2020, Big Sky, MT), Institute of Electrical and Electronics Engineers (IEEE), Mar 2020.

The definitive version is available at https://doi.org/10.1109/AERO47225.2020.9172721

Meeting Name

IEEE Aerospace Conference, AeroConf 2020 (2020: Mar. 7-14, Big Sky, MT)

Department(s)

Mechanical and Aerospace Engineering

International Standard Book Number (ISBN)

978-172812734-7

International Standard Serial Number (ISSN)

1095-323X

Document Type

Article - Conference proceedings

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2020 Institute of Electrical and Electronics Engineers (IEEE), All rights reserved.

Publication Date

14 Mar 2020