By Chris Forrester
Dr Dan Ceperley, founder and COO at LeoLabs, delivered a superb Tech Brief at the MilSat Symposium, part of the Silicon Valley Space Week, and admitted that AI seemed to be the hottest of hot topics. He said that while there seemed to be no shortage of investment cash it was worth looking at where AI was operationally extremely useful.
He looked at a couple of very useful examples. LeoLabs operates a series of radars around the world to monitor satellites and debris around the clock. “This gives us a huge data set to mine and we put a lot of effort into generating automated analysis of this data. AI enables us to take this up to another level and to augment what all the human analysts are doing.”
“But the threat landscape is changing dramatically. Satellite launches have grown every year and the situation is getting more an more congested. LEOs are getting more nimble. We are seeing proximity operations, highly manoeuvrable satellites and all this makes keeping track of what’s happening that much harder.”
He explained that three threat events and in particular the notorious Russian ‘Zombie’ satellite that woke up in 2022 after years of inactivity. Another was the Russian ‘nesting doll’ satellite which spawned multiple sub-satellites in late 2023. The third example was a set of three Chinese satellites which has been practicing various proximity operations this year. Moreover, they are not doing this work in some isolated part of space but in a highly-trafficked region.
“The fact is that it is very easy for satellites to blend in, to hide. In 2019 there were 900 satellites operating in LEO. Today it is more than 9000, a ten-fold increase in just 5 years. We have to keep track of a very large, and growing, number of satellites and to identify what is a strange activity.”
Keeping track of 50 or 100 craft is relatively easy. Today’s numbers make human observation near-impossible and LeoLabs uses a variety of deep-learning and other algorithms to handle these tasks.
He showed delegates a fascinating chart of unusual satellite activity from two days previous which had helped identify patterns of activity and which could then be examined by human eyes for more investigation.
He added that AI was useful in identifying differences between what is routine and what is deliberately different. Is the satellite friendly, or an adversary? Is it manoeuvrable, or static? Is it raising orbit or lowering?
The other role of investigation was identifying unknown satellites or objects. From 2019 to 2023 the number of unknowns grew from 400 to 1400, and these require investigation. It is also important to monitor new launches within a few hours and not days. LeoLabs updates every 12 hours and ingest the data into its tables and maps. AI is augmenting our human analysts.