The sharp rise in the number of active satellites and debris has increased the probability of satellite collision, which is a major threat to the success of space operations. This project proposes the development of an AI-based satellite collision avoidance system that can accurately predict the possibility of satellite collision using machine learning and neural networks. The Two-Line Element (TLE) data set is used to simulate the satellite orbit using standard orbital mechanics equations. For this project we have used different predictive models to suit the needs of the different types of satellite orbits, such as Low Earth Orbit (LEO), Geostationary Earth Orbit(GEO) and finally the Polar Orbits. This is done to factor for the different orbital characteristics of the satellites. The project proposes the development of frontend using opensource library like Cesium JS, which will enable us to visualize the satellites and the avoidance feature in 3D. This system will help in increasing the situational awareness and decision-making process of the satellite operators, furthermore this will also enable students and like to play around with the system.

In conclusion, Orbit Guard shows that open-source tools and publicly accessible orbital data can be used to create a complete, ML-augmented satellite collision avoidance prototype. The system achieves all five of the Phase-I report’s goals: data collection from Celestrak, orbit propagation based on SGP4, AI-driven collision probability estimation, automated alert generation, and 3D interactive visualization. It shows the feasibility of the AI-based method for use by space agencies, satellite operators, and research institutions and offers a strong basis for a production-quality space traffic management tool. Future research can go in a few different directions. First, since Space-Track Conjunction Data Messages (CDMs) contain operator-verified close approaches and full covariance matrices, substituting actual conjunction data from CDMs for the Celestrak TLE-based training labels would greatly increase model accuracy. Second, the decision support loop would be completed by incorporating a maneuver recommendation engine the existing system identifies and measures risk but does not recommend preventative measures.

[1]A. A. I. Siddiqui, “The integration of AI technologies in space traffic management,” Bentham Direct, Open Astron. J., vol. 18, pp. 1–12, Mar. 2025. [2]S. Schaefer and M. G. Cichon, “Autonomous trajectory optimization and collision management using neural systems,” IEEE Eng. Manag. Rev., vol. 53, no. 1, pp. 88–101, Jan. 2025. [3] C. Ieracitano, N. Mammone, P. Lanza, and F. Gao, “AI for space: theories, models and applications,” Neural Comput. Appl., Springer, New York, NY, USA, vol. 37, pp. 1–22, Feb. 2025. [4]T. S. Kelso, “Frequently asked questions: Two-line element set format,” CelesTrak, Colorado Springs, CO, USA, Tech. Note, 2019. [Online]. Available: https://celestrak.org/columns/v04n03/ [5]J. L. Foster and H. S. Estes, “A parametric analysis of orbital debris collision probability and maneuver rate for space vehicles,” NASA Johnson Space Center, Houston, TX, USA, NASA Tech. Memo. 100559, Aug. 1992. [6]R. P. Patera, “General method for calculating satellite conjunction probability,” J. Guid. Control Dyn., AIAA, Reston, VA, USA, vol. 24, no. 4, pp. 716–722, Jul. 2001. [7]F. R. Hoots and R. L. Roehrich, “Models for propagation of NORAD element sets,” Spacetrack Report No. 3, U.S. Air Force Aerospace Defense Command, Colorado Springs, CO, USA, Dec. 1980. [8]NASA Orbital Debris Program Office, “Orbital debris quarterly news,” NASA Johnson Space Center, Houston, TX, USA, Tech. Rep., vol. 28, no. 1, Feb. 2024. L. Breiman, “Random forests,” Mach. Learn., Springer, New York, NY, USA, vol. 45, no. 1, pp. 5–32, Oct. 2001

© 2025 International Journal of Engineering and Techniques (IJET).

Digital Object Identifier (DOI)

IJET assigns a unique DOI to every accepted article via Zenodo for persistent linking and citation. Your article’s DOI: https://doi.org/{{doi}}

Open DOI About Zenodo DOI

Close