Future 6G systems are already starting to come into focus through research and early development, and one of the most interesting aspects is the role that artificial intelligence will play in the next generation of cellular systems.
First, a note about artificial intelligence and machine learning. As Andreas Rosler, technology manager at Rohde & Schwarz, explains, AI is a branch of computer science that focuses on building intelligent machines that mimic human cognition, decision-making, and problem-solving. Machine learning is essentially a subset of AI that allows algorithms to automatically improve their performance over time and through data and experience, without any additional programming.
We already live in an era of limited use of AI, where virtual assistants and customer service bots can respond to queries and requests based on natural language processing. AI and ML are already present to a limited extent in the 5G standard, including the Network Data Analysis Function (NWDAF), which aims to collect and make available data from various nodes in the network. is defined as of release 15, he says Roessler. Deliver automated network management and optimization of individual virtualized network functions. However, he pointed out in his recent R&S webinar that only the interface is defined for this functionality, leaving the development of specific his AI/ML models to his community of vendors. Did. Additionally, in Release 15, the NWDAF functionality is quite limited in providing information about network slice load levels and is only applicable to 5G standalone mode.
But like many aspects of 5G, NWDAF provides the building blocks for broader and more interesting use cases in subsequent releases, providing a glimpse of what the next generation of wireless technology might one day look like. can do. In Release 16, the scope of NWDAF has been expanded to provide broader analysis support. In addition to load information, it will support device mobility features, user access for specific applications, subscriber user experience, sustainability information such as battery statistics, and even wireless, Roessler explains. Access network congestion information that can be relayed to operations and maintenance teams. Over time, NWDAF becomes a valuable piece of the puzzle for predictive behavior and realistic models for fine-tuning virtual network functions, mobility, session management, and QoS optimization. NWDAF is also based on the use of AI/ML.
“Network data analytics capabilities are a way to better control and improve performance of your network through automation,” says Roessler.
Meanwhile, initial discussions of the Release 18 specification include the use of AI and ML to improve the performance of the 5G New Radio air interface such as beam management, reducing channel state information (CSI) feedback overhead, and improving position accuracy. contained. There are various scenarios, he explains.
When it comes to ongoing efforts around 6G, AI and ML are key foundational technologies for many aspects of future wireless systems.
“This is not a separate field of research, but is connected to all other fields,” Rosler explains. Ultimately, AI/ML will underpin some of the features that make his 6G revolutionary. Examples from current research areas include the use of ML models to remove self-interference to enable full-duplex operation, but require significant complexity (and cost) to operate. This was difficult to achieve. AI/ML may finally be able to do just that. Additionally, regarding the 6G physical layer, academia and leading researchers are discussing how ML-based models can be used in baseband signal processing to help wireless receivers detect channel conditions and recover signal information more accurately. Rosler said many industry players are already considering it. And efficiently.
While early ML research focused primarily on the receiver side, “the next big step is to use machine learning to jointly optimize the entire transmit, receive, and baseband signal processing chain. ,” he adds.
“The ultimate goal of machine learning is to adapt the environment—the underlying hardware, signal processing techniques, and applications—to the environment,” Rosler says. So the AI/ML will actually be designing part of the 6G physical layer itself. Although it is impossible to predict exactly how it will unfold, Rohde & Schwarz believes the most likely path forward is his three-step approach. The first stage will incorporate his ML into the transceiver/RF front end and antenna system, followed by his second stage. His phase of ML integration within baseband signal processing from a receiver perspective and his third phase of end-to-end optimization.
Still, there are significant challenges and questions that must be answered in the coming years. Is AI/ML possible even in parts of wireless systems with extreme hardware constraints? Will AI systems work well if the signal contains only a few bits of information? However, one major challenge is that existing data sets may not be sufficient or may not be accessible to advance the application of AI and training of ML algorithms, Roessler said. points out.
Interestingly, the basic research on AI-native air interfaces already dates back to a 2020 paper, Roessler says. “The most interesting thing for me is that the later paper referenced the earlier paper and showed further improvements compared to the earlier findings. This clearly shows that this methodology has great potential. Rosler added. Learn more about the potential applications of AI/ML in 6G systems and how Rohde & Schwarz is supporting early research and development.