New AI Model Enhances Space Safety By Monitoring Satellites

A groundbreaking AI model is set to revolutionise space safety by accurately predicting satellite anomalies and preventing collisions, crucial for managing the rapidly growing number of objects in Earth's orbit.

A new AI model can monitor satellites and spot anomalies in their behaviour, reducing the risk of in-orbit collisions, according to new research published in the journal Expert Systems. The work was carried out as part of AI4 Space Safety and Sustainability, a UK Space Agency International Bilateral fund consortium across FVEY countries â Australia, Canada, New Zealand, the United Kingdom, and the United States. The project is led by Professor Massimiliano Vasile, director of the Aerospace Centre of Excellence at the University of Strathclyde, the Alan Turing Institute, the University of Arizona, MIT, the University of Waterloo, and industry partners GMV, Columbiad, LMO, and Zendir.

How The AI Model Works

The AI model, developed by researchers from the Alan Turing Institute's Defence AI Research Centre (DARe), is the first to predict anomalies and satellite motion by learning from the way light is reflected off objects in space. It could help to address the growing challenge of safeguarding thousands of satellites by detecting objects in orbit to aid space traffic management and inform collision avoidance manoeuvres.

The world is increasingly reliant on satellites, but space is becoming increasingly crowded, with more than 4,000 new satellites launched in 2025, compared to just 159 in 2000. Monitoring these is challenging and time-consuming, but the fully automated AI tool can detect satellites behaving unusually, predict their motion through space, and forecast future behaviour.

Training And Real-Time Application

The tool has been trained on huge quantities of satellite brightness readings or "light curves" gathered through telescopes to understand normal patterns of satellite behaviour. After this training, the model is 'fine-tuned' with highly curated simulation data from both the Aerospace Centre of Excellence at the University of Strathclyde and GMV.

Once in action, the model is fed real-time or recent light curves from ground-based observatories and then it flags anomalies so human experts can investigate further.

Impact And Future Developments

Testing has shown that the tool can identify unusual or unexpected light curves 88 per cent of the time and is also able to distinguish between different behaviours of a satellite, such as spinning versus tumbling, analyses which are essential for in-orbit servicing and extending the life of satellites. It demonstrates the potential for systems like this to enable real-time anomaly identification and would allow human operators to investigate automatically flagged issues quickly and take action to avoid collisions.

Professor Massimiliano Vasile, Professor of Space Systems Engineering and author and lead of the AI4S3 project, said: "With AI4S3 we demonstrated that AI can make a difference in multiple areas of space safety. Space object behavioural analysis is one of those areas." He said that understanding and explaining the behaviour of space objects is critical to predict the evolution of the whole space environment and guarantee the safety of essential services for everyday life. "In the Aerospace Centre we have been working on understanding the motion of space objects for a long time but with AI4S3 I wanted to see if modern AI technology could help to detect regular and anomalous behaviours even from a single pixel in the sky," Vasile said. He said that the results obtained by the Alan Turing Institute are indeed remarkable and a key step towards a complete and systematic analysis of the behaviour of any resident space object.

Next steps for the project include researching the potential for multi-modal systems to also include radar data, hyperspectral data and satellite orbit data, which could provide even greater insights to experts monitoring and safeguarding satellites.