“It was an ideal time,” says Petar Popovski about the beginning of his work as a U Bremen Excellence Chair holder. At the start of 2019, the topics of satellite communications and non-terrestrial networks (NTNs) had just entered the agenda of the 3rd Generation Partnership Project (3GPP), the main standardization body for mobile communications worldwide. Popovski and his Bremen colleague Professor Armin Dekorsy from the Department of Communications Engineering at the University of Bremen quickly realized that they could carry out pioneering work with their research project, as the topic would gain significant momentum in the following years. And indeed: seven years later, NTNs have become a major focus in research, standardization, and commercial development in satellite communications.

How Do Satellites Learn?

Petar Popovski is one of the world’s leading experts in wireless communication. This highly decorated electrical engineer conducts research and teaches at Aalborg University in Denmark. Since 2019, he has been one of eleven U Bremen Excellence Chair holders. These are top international researchers who act as a bridge between the University of Bremen and other leading universities worldwide. Popovski’s connections to Bremen date back to 2012. At that time, his team worked with Armin Dekorsy’s research group on projects that helped lay the foundations for 5G technology. Building on the insights and methods developed during that period, the team has advanced its research on NTNs within the Distributed Learning in Space project. Their work focuses on making satellites capable of learning using AI and sharing their knowledge with one another.

Together with Dekorsy, postdoctoral researcher Dr. Bho Matthiessen (now a professor), and PhD student Nasrin Razmi, Popovski addressed the challenge of federated learning in satellite systems at the very beginning of the project. “We saw this as a unique opportunity to define a meaningful research direction at the intersection of satellite communications, AI, and communication theory, since there was virtually no prior work in this area,” he recalls. “Today, this has become a vibrant research topic within NTNs, and our early contributions have received significant attention in terms of downloads and citations.”

Large Volumes of Data, Limited Transmission Speed

The most challenging aspect, Popovski says, was “defining a system model that accurately captures the operational constraints and requirements of satellite networks.” Satellites collect vast amounts of measurement data about the Earth while in orbit. This data is used not only for weather forecasting and navigation, but can also serve as an early warning system for natural disasters such as wildfires, landslides, and floods. If satellites learn to share relevant data among themselves and transmit prioritized information to Earth, they can be used more effectively. This is particularly important given the large volume of data generated by Low Earth Orbit (LEO) satellites, which are the focus of Popovski and Dekorsy’s research. These are small satellites that orbit the Earth at an enormous speed of 28,000 kilometers per hour – nearly eight kilometers per second. They typically operate at altitudes between about 250 and 2,000 kilometers, making them close enough to Earth to deliver large quantities of high-resolution images and meaningful measurement data. “Not all the data collected by the satellites needs to be transferred to the ground,” says Popovski. At the same time, the non-transmitted data remain valuable because they serve as important material for learning and inference within the satellite constellation. “This, of course, requires elaborate coordination and communication with the terrestrial facilities, for example, in terms of tuning the AI models used by the satellites.”

Improving System Efficiency

AI plays a central role in the development of NTNs. “Having AI in the satellite constellations will be instrumental in processing and storing the relevant data collected by the satellites,” explains Popovski. “Even more, it will facilitate the communication with the terminals and stations on the ground, as AI can be used to orchestrate the communication resources.” Data rates are more limited in the uplink direction – from user equipment on Earth to the satellite. AI can help to manage this imbalance by prioritizing which data should be transmitted and when, Popovski emphasizes. This area of satellite communication is known as spectrum management and focuses on using available radio resources as efficiently as possible, for example when a software update needs to be installed. “This improves overall system efficiency.” It is a topic that Popovski and Dekorsy will continue to explore in the second phase of the project. The sixth generation of mobile communications, 6G, with its commercial deployment in the coming years, could prove helpful here. In addition to smartphones, other components such as drones, aircraft, and satellites could operate within the same communication standard.

When asked about his collaboration with Petar Popovski, Armin Dekorsy immediately lists several positive attributes: “Inspiring, creative, stimulating, open, goal-oriented, and very trusting.” The collaboration has also had a very positive impact on the support of early-career researchers at the institute. “Both his scientific impulses and ideas, as well as becoming familiar with his methodological approach to research questions, provide significant added value for my staff and contribute substantially to their academic development,” says Dekorsy, who heads the Department of Communications Engineering at the University of Bremen.

Successfully Acquiring Third-party Funding

The two researchers also benefit from their distinct but complementary networks. While Popovski focuses primarily on fundamental research, Dekorsy’s work is more technology and knowledge transfer-oriented. “This combination increases the opportunities to participate in collaborative international research and third-party funded projects,” Dekorsy says. This approach has already proven successful in the first phase of the project. The team successfully acquired funding from the EU, the European Space Agency (ESA), and the German Research Foundation (DFG).

In the second phase of the project, one of the goals is to transfer the findings into industry. “Both Bremen and Aalborg host strong clusters of satellite-related industries and renowned research groups at their universities,” says Popovski. He and Dekorsy are therefore planning a meeting between the two clusters. “We believe that this will lay the foundation for even closer collaboration between the industries and the two universities.”