University of Luxembourg: Improving satellite communication with the power of AI

The presence of artificial intelligence (AI) is becoming increasingly commonplace, making small improvements in almost every facet of life—from health care to finance, transportation and more. As research and innovation in this domain continues to flourish, a dynamic team of SnT researchers, led by Dr. Eva Lagunas, is reaching critical milestones in their journey to uncover the next frontier of AI research: satellite communications. As satellite service providers aim to bring wireless communications to the most remote areas of the planet and even beyond, the SmartSpace initiative is looking specifically at how AI can help.  

Over recent decades, the structure of satellite communications has evolved from single geostationary (terrestrial) satellites to a (non-terrestrial) cluster or swarm of small satellites flying in perfectly coordinated orbits. At present, the coordination and control of satellite resources is contingent on time-varying conditions such as satellite movement and wireless channels. Yet existing model-based techniques that allow non-terrestrial satellites to perform well are computationally complex and time-intensive. Using machine learning, Lagunas and her SmartSpace team have found new ways that AI can learn to mimic existing techniques and more rapidly compute the best network configuration to deliver services more efficiently. For the everyday user, the research will help mitigate service interruptions caused by poor signal quality and traffic congestion. AI will help predict these situations and generate preemptive solutions to avert potential link outage or disruption.  

Partnership with a telecommunications giant

At the centre of SmartSpace is an industry partnership with leading telecommunications giant, SES. The benefits for SES are clear: the company wants a deep investigation into new technologies that will translate into a competitive advantage. For researchers, having SES as industrial advisor in the project allows a clearer lens into the day-to-day practical problems faced by satellite operators. As Lagunas explains, “when my team can see first-hand the barriers faced by a satellite operator, we can deploy our tools with greater synchronicity to achieve targeted gains. It is exciting for us as researchers to show how AI can solve problems—how it can be meaningfully and contribute to tangible performance gains.”  

Yet despite the clear benefits, some remain hesitant of AI. “It is true that AI does not fit all businesses,” Lagunas explains. AI and machine learning requires large amounts of data, which needs to be collected, refined, and verified as representative of the targeted goal. This can make many companies—especially small-to-medium sized businesses—concerned about capacity. “But SmartSpace is ready to work with telecommunications and space actors of all varieties to demonstrate what the technology can do for different use-cases and scenarios. For example, we can shed light into the optimization of storage facilities by pointing out relevant data that needs to be collected or we can provide guidance on appropriate post-processing techniques that can orient data collection to align with certain objectives. There’s a lot we can do and we want people to know we are here and ready to help.” The team has synthetic datasets obtained from internally developed software that are freely available online.  

Now in its second year, SmartSpace also provides a fertile training ground at SnT. As global leaders in their field, the team is attracting top talent from around the world. For Almoatssimbillah Saifaldawla, who is currently completing a PhD, the decision to join The University of Luxembourg was easy. “With both satellite communications and AI evolving as quickly as they are, I wanted to work with the best team to help explore these futuristic technologies.” Saifaldawla received praise for his research on interference detection at recent top conferences in Canada and the United States. Lagunas is proud of such achievements: “We have a very strong team of research associates. Each is investigating a different aspect of the project such as network load prediction, precoding calculation, research management and their research is already available in the public domain. With our recent acquisition of hardware for neuromorphic computing, I expect to continue see impressive results from the team as our project evolves in coming years.” 

This article was first published on 4 September by University of Luxembourg.