“Perhaps the most important thing is that we have now shown that it is possible to group the frequency resources better with the use of machine learning, for mobile communication in the future can be faster and better and support the new digital future reality 6G opens to.”

So says Rebekka Olsson Omslandseter at the Department of ICT at the University of Agder, who recently submitted her doctoral thesis "On the Theory and Applications of Hierarchical Learning Automata and Object Migration Automata".

In her work, Omslandseter has looked at the challenges faced by frequencies in future international mobile communication networks.

The backdrop is 6G - the sixth-generation technology standard for wireless communication. This is still under development but is expected to take over for today's 4 and 5G standards by 2030.

The 6G standard will be significantly faster than previous generations. It will also support other applications. Among these is the Internet of Things, where virtually all electrical appliances can be connected to the Internet, including self-driving cars and refrigerators that tell you what to refill when needed.

In such a future, with ubiquitous connectivity between billions of devices – not to mention an infinite array of sensory experiences such as games and other forms of entertainment – 6G requires new frequency spectrum considerations. The frequency bandwidth used in the 4G and 5G network today is simply not large enough for tomorrow's needs.

One of the ways to solve the challenges is to expand the width of the frequency band beyond what is used today. Another is to make better use of the activity in the frequency area. And this is where Rebekka Olsson Omslandseter’ research comes into its own.

In her work, she focuses on expanding the problem-solving ability and improving the efficiency of learning machines. Especially related to Object Migration Automata (OMA), which is a type of machine learning algorithm.

“OMA algorithms are particularly interesting because they are systems within artificial intelligence that can organize and reorganize various objects even if their situations change. They do it on the fly, so to speak”, she explains.

The "objects" in this context can be anything from files in a database, to animals on a farm, or goods in an online store - or users and activity in a mobile network.

“Imagine a system that can effectively group and manage the frequency resources in a mobile network, and thus always ensure the best possible and most efficient connection for everyone - then you have a system where an OMA algorithm plays centre stage”, she says.

The algorithm Omslandseter proposes in the next generation mobile network can therefore be closely linked to and adapt to the activity of the users. This also applies to those who behave extremely randomly. This is something that has not been highlighted in previous research.

In addition, Omslandseter's research has demonstrated how AI learning machines can learn both faster and better, and thus become able to solve problems they could not solve before.

Central to this is a special learning structuring which significantly reduces the calculation work needed. Her research shows that it is possible, with the help of machine learning and learning units, to reduce the number of calculations by 40 percent; in a given example from 1 million to 600,000.

“All processing of data requires a lot of resources, energy. When developing new methods, it is therefore important that the footprint of data processing is reduced. It has been good to be able to contribute to this too, through my work”, says Rebekka Olsson Omslandseter.

In continuation of her doctoral work, the Omlandseter is now working on developing a system that continuously and correspondingly learns and adapts to protect the devices of the Internet of Things against cyber threats. Here, too, the OMA algorithms are central.