Could AI and Machine Learning revolutionize real-time space situational awareness? By Heer Patel

Orbital space was once a wide-open expanse of nothingness, but it is now becoming significantly blocked up, with thousands of satellites orbiting around the Earth. This has led to a relentless increase in space debris, and the need for real-time Space Situational Awareness (SSA) has been more crucial than ever. The current traditional methods cannot keep up with the exponential growth of potential obstacles that pose a risk for collision, but Artificial Intelligence (AI) and Machine Learning (ML) are being developed as potential solutions, which will assist in revolutionizing how we humans manage and interact with our orbital environment. 

The heart of this revolution is the ability of AI to process and compute substantial amounts of data. Ground-based sensors, satellite telemetry, and space-based telescopes tend to generate information that can be overwhelming for humans to understand thoroughly without assistance from AI or ML. According to MarketsandMarkets and Number Analytics, AI or ML would critically improve the accuracy of object detection, fusing data from various sensors to generate an understandable and frequently updated orbital map. This capability assists with reducing false potential and in providing early warnings to humans to prevent collisions or severe failures that could pose a threat to man-made objects.  

Furthermore, AI excels in capabilities that are paramount in active threat mitigation. The large number of objects in Earth’s orbit results in collision avoidance no longer being a measure that could be prevented by reaction-based maneuvers. Deep learning models can be used in empowering satellites with the ability to perform orbital maneuvers independently and continually simulate various scenarios to determine optimal paths, which can minimize fuel consumption and reduce the need for continual ground control. Additionally, Number Analytics discusses how AI holds the ability to analyze significant amounts of data from sensors, historical collision data, and orbital predictions to allow for proactive measures, as both NASA and the ESA are currently attempting to integrate AI for collision avoidance.    

AI is rapidly becoming the backbone of new Space Traffic Management solutions, which allows for more dynamic traffic control via the prediction of movement patterns and potential clustered areas, such as in Low Earth Orbit. AI-based platforms can coordinate satellite movements through various orbital trajectories and altitudes, while reducing conflict risk and fostering cooperation with ground operations, according to MarketsandMarkets and KeAi Publishing discusses research in integrating learning, game theory, and optimal control to improve the autonomy in satellites to reduce potential collision risks. Branching from AI satellite integration, KeAi Publishing discussed supporting a fully autonomous satellite management system, which would additionally be decentralized. 

AI could also be used in reducing the continually growing space debris, such as in identifying debris using radar signals and optical imagery, to then be used in modeling fragmentation events from debris collisions. Research in such technology presented in “Machine learning with oversampling for space debris classification based on radar cross section” by Y. Zhang, showcases fusing Support Vector Machines via oversampling methods achieving 99% accuracy in determining potential hazardous debris from low-risk debris, even with being trained on incorrect datasheets. This level of precision assists in prioritizing the tracking of high-threat objects and in debris removal missions. 

Current advancements in low-power AI-oriented computing hardware also allow for AI integration directly on spacecraft. M. Lim et al., in their 2020 AMOS Conference paper called “Onboard Artificial Intelligence for Space Situational Awareness with Low-Power GPUs,” explore how such technology can be used towards reducing latency, which allows for real-time processing and immediate response without the need to downlink all unfiltered data to ground control. As highlighted by KeAi Publishing, AI can provide high-precision dynamics modeling and AI-enhanced orbital logging to assist with navigation. 

Furthermore, machine learning tends to heavily improve the accuracy of satellite orbit prediction. H. Peng and X. Bai, within their work, which was published in The Journal of the Astronautical Sciences, often showcase how ML models such as Support Vector Machines and Artificial Neural Networks can learn from previous historical orbit prediction errors via the usage of publicly available data. This will improve the precision physics-based models, which have been used traditionally. This capability is additionally essential in managing the large number of resident space objects and in reducing the risk of large collisions. 

At EcoAero, we recognize that managing orbital space is rapidly shifting from a question of the future to a responsibility of the present. Traffic collisions and congestions in space are issues that only grow more pressing as space debris counts increase. Using AI and machine learning in space situational awareness could lead to a safer and more sustainable future. We believe that innovation and advances in technology should be embraced, rather than feared, and integrated, rather than sidelined. The use of AI and machine learning in space situational awareness could help clear the way for the future of space travel.

Image courtesy of the European Space Agency. Used for editorial purposes only. No commercial use.