Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{205050,
        author = {Subiya Begum and Dr. B Ramesh and Ramya S L and Nida Falak},
        title = {An AI-Based Space Debris Detection and Avoidance System: A Review Paper},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {1},
        pages = {5097-5106},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=205050},
        abstract = {The steady accumulation of orbital debris around Earth represents an escalating hazard to functioning spacecraft, demanding autonomous on-board systems capable of identifying imminent threats while conserving the finite propellant reserves that govern satellite longevity. This paper proposes a machine-learning-driven architecture for simultaneous debris detection and fuel-conscious collision avoidance, developed entirely in Python. The framework maintains a dynamic simulation of one protected satellite alongside several debris’ bodies, propagating all trajectories through fourth-order Runge-Kutta integration. A Random Forest classifier ingests five relative-motion features separation distance, relative speed, approach angle, distance rate-of-change, and cross-track angular momentum to estimate the probability that a tracked object will enter a 500-metre exclusion zone within a 60-second lookahead window. When the classifier reports a probability above 0.7, or when a debris body is already inside the exclusion boundary, an optimisation engine engages. That engine systematically searches six impulsive thrust directions (radial, in-track, and cross-track in both senses), selecting whichever requires the least delta-v to widen the predicted closest approach to at least 200 metres. Across a 10,000-encounter validation set, the classifier attained 94.2% accuracy and an AUC-ROC of 97.1%, while the optimisation routine reduced mean propellant expenditure by 37.8% relative to a fixed radial-burn baseline. Per-object inference completes in under 5 ms, consistent with deployment on small-satellite flight computers. Matplotlib-based animations provide concurrent visual confirmation of spacecraft and debris motion. The resulting system constitutes a scalable, computationally lean pathway toward fully autonomous satellite protection in progressively congested near-Earth orbital regimes.},
        keywords = {Collision Avoidance, Machine Learning, Orbital Debris, Random Forest, Satellite Safety, Space Debris},
        month = {June},
        }